#Mitin - Introduction To Nano Electronics

Introduction to Nanoelectronics
Science, Nanotechnology, Engineering,

and Applications
VLADIMIR V. MITIN
University of BufTalo, State University of New York
VIATCHESLAV A. KOCHELAP
Institute of Semiconductor Physics National Academy of Sciences, Ukraine
MICHAEL A. STROSCIO
University of Illinois at Chicago
UCAMBRIDGE
::" UNIVERSITY "KESS
Introduction to Nanoelectronics
Increasing miniaturization of devices, components, and integrated systems requires

developments in the capacity to measure, organize, and manipulate matter at the
nalloscale. This textbook is a comprehensive, interdisciplinary account of the technology
and science that underpin nanoelectronics, covering the underlying physics, nanostruc·
lures, naoomaterials, and nanodcviccs.
Without assuming prior knowledge of quantum physics, this book provides a unifying
framework for the basic ideas needed to understand the recent developments in the field.
Following an introductory description of recent trends in semiconductor and device
nanotcchnologies, as well as novel device concepts, materials for nanoelcctronics arc
treated, covering methods of growth, rabrication and characterization. Treatment then
moves to an analysis of nanostructures, including recently discovered nanoobjects, and
concludes with a discussion of devices that use a "simple" scaling-down approach to
copy well-known microelectronic devices, and nanodevices based on new principles that
cannot be realized at the macroscale.
Numerous illustrations, homework problems and interactive Java applets help the
student to appreciate the basic principles of nanotechnology, and to apply them to real
problcms. Written in a clear yet rigorous and interdisciplinary manner, this textbook is
suitable for advanced undergraduate and graduate studcnts in electrical and electronic
cngineering, nanosciencc, materials, bioengineering, and chemical engineering.
Further resources for this title, including instructor solutions and Java applets, are
available online at www.cambridge.org/9780521881722.
Vladimir V. Mitin is a Professor and Chair of the Department of Electrical Engineering

at the University of Buffalo, State University of New York. He has co-authored eight
books, and over 400 professional publications, including ten patents.
Viatcheslav A. Kochelap is Professor and Head of the Theorctical Physics Departmcnt

at the Institutc of Semiconductor Physics ational Acadcmy of Sciences, Ukrainc. He
has published over 200 journal articles.
Michael A. Stroscio is a Professor in the departments of Electrical and Computer Engi-

neering, and Bioengineering, at the University of Illinois at Chicago. He is a Fellow of
the IEEE, the American Physical Society, and the AAAS.
CAMBRIDGE UNIVERSITI PRESS
Cambridge. ew York, Melbourne. Madrid, Cape TOWIl, Singapore, Sao Paulo
Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United Srares of America by Cambridge Universiry Press, New York
w\vw.cambridge.org
Informacion on this ride: www.cambridgc.org/9780521881722
© Cambridge University Press 2008
This publication is in copyrighr. Subject to Sr3[U[ory exception and (0 the provision of

relevant collective licensing agreements, no reproduction of any parr may take place
withour the written permission of Cambridge University Press.
First published in prill{ [ormar 2007
ISBN-I3 978-0-51 1-36964-3 eBook (ebra'l)

ISBN-IO 0-511-36964-6 eBook (rbra'l)
ISBN-13 978-0-521-88172-2 hardback

ISBN-IO 0-521-88172-2 hardback
Cambridge Universiry Press has no responsibility for the persistence or accuracy of uris
for external or third-parry internet websites referred to in this publication, and does not
~uaranrcc that any coment on such websites is, or will remain, accurate or appropriate.
Preface
Welcome to the amazing nallolVorldl In this book you will find fundamental princi-
ples in nanoscience and basic techniques of measurement, as well as fabrication and
manipulation of matter at the nallosenIc. The book discusses how these principles, tech-
niques, and technologies arc applied 10 the newest generation of electronics, known as
nonoelect1Vnics.
The science of atoms and simple molecules, and the science ofmatter from microstruc·
tures to larger scales, are both well established. A remaining, extremely important, size·
related challenge is at the nanoscale -roughly the dimensional scales between 10 and 100
molecular diameters - where the fundamental properties of materials are determined and
can be engineered. This field of science - nanoscience - is a broad and interdisciplinary
field of emerging research and development.
Nanotechnology is concerned with materials. structures, and systems whose compo-
nents exhibit novel and significantly modified physical, chemical, and biological prop-
erties due to their nanoscale sizes. A principal goal of nanotechnology is to control and
exploit these properties in structures and devices at atomic, molecular, and supramolecu-
lar levels. To realize this goal, it is essential to learn how to fabricate and use these devices
efficiently. Nanotechnology has enjoyed explosive gro\\1h in the past few years. In partic-
ular, nanofabrication techniques have advanced tremendously in recent years. Obviously,
revolutionary changes in the ability to measure, organize, and manipulate matter on the
nanoscale are highly beneficial for electronics with it's persistent trend of downscaling
devices, components, and integrated systems. In turn, the miniaturization required by
electronics is one of the major driving forces for nanosciencc and nanotechnology.
Practical implementations ofnanoscience and nanotechnology have great importance.
and they depend critically on training people in these fields. Thus. modern education
needs to address the rapidly evolving facets of nanoscience and nanotechnology. A new
generation of researchers, technologists, and engineers has to be trained in the emerg-
ing nanodisciplines. With the purpose of contributing to education in the nanofields,
we present this textbook providing a unifying framework for the basic ideas needed to
understand recent developments underlying nanoscience and nanotechnology, as applied
to nanoelectronics. The book grew out of tile authors' research and teaching experience
in these subjects. We have found that many of the ideas and achievements in fields under-
lying nanoscienee and nanotechnology can be explained in a relatively simple setting,
if the necessary foundational underpinnings are presented properly. We have designed
this textbook mainly for undergraduate students, who will be trained in diverse fields
viii Preface
including nanoscience, physics of material devices, electrical and optical engineering,

materials science and engineering, and mechanical engineering. It can be helpful also
for training students in bioengineering and chemical engineering. To reach such a broad
audience, materials are presented in such a way that an instructor can choose the level
of presentation depending on the backgrounds of the students. For example, we have
included Chapters 2 and 3 in part for students who have not taken a quantum mechanics
course. An analogy with wave fields - elastic waves and optical waves - is exploited
widely to introduce wave mechanics of particles and quantum principles, which play key
roles in the interpretation of the properties of nanomaterials.
One of us (VV.M.) has taught the course for students in the second semester of their
sophomore year. For students at this level, Chapters 2 and 3 were covercd in detail and,
consequently, there was not enough timc to cover all of the devices that arc discussed in
Chapter 8. If students using the book have previously taken courses on quantum mechan-
ics and clectromagnctics, the instructor may start from Chapter 4. This book may be also
lIsed as an introductory graduate or senior undergraduate course. Another of us (M.A.S.)
has used Chapters 2 and 3 as the introduction to a graduate course on nanoclectronics for
a class with students drawn from electrical engineering, materials engineering, chem-
ical engineering, mechanical engineering, and physics. By covering Chapters 2 and 3
at the beginning of the course, the students can then proceed from this common basis
in quantum mechanics and Olher underlying areas of physics to cover more advanced
topics, either in the current text or in other texts such as QUlin/lim Heterostl1lctures by
V Mitin, V. Kochelap, and M. Stroscio. The latter approach has been used by M.A.S. in
teaching nanoelectronics to graduate students \\~th diverse backgrounds in many disci-
plines within engineering and the physical sciences. For this purpose, we include details
of derivations and mathematical justification of concepts in some sections. These details
can be omitted from an undergraduate curriculum.
The book contains homework problems on various subjects. These problems illustrate
the basic material and help students to understand and learn the basic principles of the
nanoscience and the nanotechnology.
.. .... .
Essentially, the chapters are organized into three main groups.
Chapters 1-3 are of an introductory character. In Chapter I, we present in concise form
the main subject of the book. The recent and diverse trends in semiconductor and device
nanotechnologies, as well as novel concepts of nanodevices. are reviewed. These trends
make it clear why understanding the fundamentals ofnanoscience is of great importance.
Chapters 2 and 3 are written for students who have not taken a quantum mechanics
course. An analogy with wave fields (elastic waves and optical waves) is exploited widely
to introduce wave mechanics of particles and the quantum principles, which play key
roles in the interpretation of the properties of nanomaterials.
In Chapter 2, we explain Ihal the fundamental laws of physics governing particles
and material fields in the nanoworld are different from those that apply to familiar
macroscopic phenomena. Starting with an analysis of an arbitrary wave field (clas-
tic vibrations in solids, electromagnetic fields, etc.), we find particlc-Iike behavior of
this wave field for small wave amplitudes and (or) for spatial scales larger than the
1 Toward the nanoscale
This book provides the foundations and the main ideas emerging from research that
underlies the applied field called IIalloelectronics. Nanoelectronics promises to improve,
amplify, and partially substitute for the well-known field of microelectronics. The prefix
micro denotes one ",illionth and, as applied to electronics, it is used to indicate that
the characteristic sizes of the smallest features ofa conventional electronic device have
length scales of approximately a micrometer. The prefix JlallO denotes olle billiomh.
Thus, in nanoeleclronics the dimensions of the devices should be as many as a thousand
times smaller than those of microelectronics.
uch a revolutionary advance toward miniaturization of electronics is based on the
recently developed ability to measure, manipulate, and organize matter on the nanoscale
- I 10 100 nanometers, i.e., I to 100 billionths of a meter. At the nanoscale, physics,
chemistry. biology, materials science, and engineering converge toward Ihe same princi-
ples and tools, and form new and broad branches of science and technology that can be
called lU/1Iosciellce and nanofecllllology.
Advancing to the nanoscale is not just a step toward miniaturization, but requires the
introduction and consideration of many additional phenomena. At the nanoscale, most
phenomena and processes are dominated by quantum physics and they exhibit unique
behavior. Fundamental scientific advances are expected to be achieved as knowledge in
nanoscience increases. In turn, this will lead to dramatic changes in the ways materials,
devices, and systems are understood and created. Innovative nanoscale properties and
functions will be achieved through the control of matter at the level of its building blocks:
atom-by-atom, molecule-by-molecule, and nanostructure-by-nanostructure. The molec-
ular building blocks of life - proteins, nucleic acids, carbohydrates - are examples of
materials that possess impressive properties determined by their size, geometrical fold-
ing. and patterns at the nanoscale. Nanotechnology includes the integration of manmade
nanostructures into larger material components and systems. Importantly, within these
larger-scale systems, the active elements of the system will remain at the nanoscale.
The driving forces underlying developments at the nanoscale have at least two major
complementary components - scientific opportunities and technological motivations.
Scientific opportunities
The progress in physics, chemistry, and biology at the nanoseale represents a natural step
in advancing knowledge and understanding Nature. Scientific perspectives on this route
are conditioned first of all by new quantum phenomena in atomic- and molecular-scale
structures and by the interaction of large numbers of these small objects. Indeed, the
fundamental laws of physics in the nanoworld differ from those that apply to familiar
macroscopic phenomena. Instead of classical physics, that works so well for macroscopic
phenomcna. the motion of particles and systems in the nanoworld is dctermined by the
so-called wavc mechanics or quantum mechanics. A basic principle of nanophysics is the
fundamental concept that all matter, including electrons, nuclei, atoms, electromagnetic
fields, elc., behaves as bOlh waves and particles. This wave-particle duality of all matter
is strikingly apparent at the nanoscale. For dealing with a large number of particles or
systems, the statistical laws are important. Statistical physics on the nanoscale is also
fundamentally different from that on the macroscale. In general. phenomcna that involvc
very large numbers of small interacting particles or systems follow different rules from
those involving only a few of them. Cooperative behavior of many-object systems is
rcvealed clearly at the nanoscale. Besides the phenomena just discussed, there arc other
classes of phenomena that arc important for science at the nanoscalc.
It is appropriate here 10 refer to the famous 1959 lecture of the Nobel Prize laureate
Professor Richard Feynman wilh the title "There is plenty of room at the bottom," where
he discussed "the problem of manipulation and controlling things on a sillall scale."
Feynman did not just indicate that there is "room at the bottom," in terms of decreasing
the sizc of things. but also emphasized that there is "plenty of room." In his lecture,
Feynman justified the inevitable development of concepts and technologies underly-
ing the nanoworld and presented his vision of exciting new discoveries and scientific
perspectives at the nanoscale.
Technological motivations
Achievements in nanoscience and nanotechnology will have tremendous multidisci-
plinary impacl. The benefits brought by novel nanotechnologics are expected for many
important practical fields of endeavor. These include materials and manufacturing, elec-
tronics, computcrs, telecommunication and information technologies, medicinc and
health. the environment and energy storage, chemical and biological technologies. and
agricullllre. Having stated the purpose of this text. wc consider now more detailed moti-
vations for the development of electronics at the nanoscale.
In general. progress in electronics is stimulated, in part, by the enormous demands for
information and communication technologies as weU as by the development of numer-
ous special applications. The continuous demands for steady growth in memory and
computational capabilities and for increasing processing and transmission speeds of
signals appear to be insatiable. These determine the dominant trends of contemporary
microelectronics and optoelectronics. Onc of the main trends of tile progress in electron-
ics was formulated by Intel co-founder Dr. Gordon Moore as the following empirical
obscrvation: the complexity oj integrated circuits. with respect 10 minimum component
cost, doubles eve,:)! 24 II/0I11"S. This statement formulated forty years ago is known as
Moore's law and provides an estimate of the rate of progress in the electronics industry.
Specifically, Moore's law predicts that the number of the basic devices - transistors - on
nl"FI~olc 'n''Of1~I~"''~ Jil

•..-
Toward the nanoscale 3
a microchip doubles everyone to two years. This is possible only if progressive scaling
down of all electronic components is realized.
Electronics exploits the electrical properties of solid-state materials. A simple and
intuitive classification of solids makes a distinction between dielectrics and metals, i.e.,
dielectrics are non-conducting materials whereas metals are good conducting materials.
Semiconductors occupy the place in between these two classes: semiconductor materials
are conducting and optically active materials with electrical and optical properties vary-
ing over a wide range. Semiconductors are the basic materials for microelectronics and
remain the principal candidates for use in nanoelectronic structures because they exhibit
great flexibility in terms of allowing the control of the electronic and optical properties
and functions of nanoelectronic devices. Accordingly, to a large extent, we will analyze
the trends of electronjcs in the context of semiconductor technology.
It is instructive to illustrate these tTends and achievements through the example of
Si-based electronics. Indeed, contemporary microelectronics is based almost entirely
on silicon technology, because of the unique properties of silicon. This semiconductor
material has high mechanical stability as well as good electrical isolation and thermal
conductivity. Furthermore, the thin and stable high-resistance oxide, SiOl, is capable
of withstanding high voltages and can be patterned and processed by numerous meth-
ods. Silicon technology also enjoys the advantage of a mature growth technology that
makes it possible to grow Si substrates (wafers) of larger areas than for other semi-
conductor materials. The high level of device integration realizable with Si-based elec-
tronics technology may be illustrated by the important integrated circuit element of any
computer, controller, etc. - the dynamic random access memory (DRAM). The main
clements of DRAM based on complementary metal-oxide-semiconductor technology
(Si-CMOS) are metal--oxide-semiconductor field-effect transistors (MOSFETs). For Si
MOSFETs, channels for flow of electric current are created in the Si substrate between
the source and drain contacts, and the currents are controlled by electrodes-metal gates-
which are isolated electrically by very thin Si02 layers. which have become thinner than
10 nm.
Figure 1.1 illustrates the evolution of the DRAM size and transistor gate size as
functions of time. Besides transistors and capacitors, the chip contains metallic line
connections: local, intermediate, and global wiring. Figure 1.1 illustrates the steady
scaling down of all characteristic sizes and increasing levels of integration. For example.
the 64-Mbit DRAM chip contains approximately 10 8 transistors per cm- 2, each with
feature sizes of the order of 0.3 \l-Ill. The transistors in this DRAM as well as those of
the more highly integrated 256-Mbit chip operate as conventional devices and obey the
laws of classical physics. The next generation ofdevices is entering the nanoscale regime
where quantum mechanics is important; indeed, as we will discuss in this book, quantum
mechanics becomes dominant on the scale of approximately one to ten nanometers for
devices that operate at room temperature. According to Fig. 1.1, loday's technology has
already reached the nanoscale and newer device concepts should be implemented before
2010.
One of the factors driving the huge production and wide use of microelectronic sys-
tems is the relatively low cost of their fabrication. Moreover, despite their increasing
nl ' , a !~~Io.,,- 'n "~~.1 "~Ic,,~

IIII
Table 1.1 Aroadmap for Si-based microelectronics (predictions of the Semiconductor

Industry Association)
1995 1998 2001 2004 2007 2010
Memories, DRAM
Bits per chip 64M 256M 1G 4G 16 G 64G
Cost per bit (miUi-ccnt) 0.017 0.007 0.003 0.001 0.0005 0.0002
Cost per chip (USS) 11 18 30 40 80 130
Logic. microprocessors
Transistors per cml 4M 7M 13 M 25 M 50M 90M
Cost pef transistor (milli-ccnt) 1 0.5 0.2 0.1 0.05 0.02
Power supply (V) 3.3 2.5 1.8 1.5 1.2 0.9
PummClers
Minimum feature size (J-lIn) 0.35 0.25 0.18 0.13 0.10 0.07
Wafer size (in.) 8 8 12 12 16 16
Electrical defect density per m 2 240 160 140 120 100 25
The data arc from U. Konig, Physica Scripta, T68, 90, 1996.
350 '--,~.~•.,.._p_1_a_na_'_C_M_O_S_D_OV_;_'. 1 Novel Device ConCQpts

".
250 ' . .... ..
- 180
• . ... ,
I
-.....
........
. ' .. '.
• ...
'.
Local Wiring
······;?::;;;in
,.~~....
-•••
E 130
c
90 ;;;;-"".i~.",:~::·-,".:=:>.......~~ .'~.;:.~ -- g
- ..
.........-'" -. -
~
'",e
-••
,E
65
45
32 i,
,•
• ,.•
•
--.,.. --.-.
.... -
.-.....
... .... ..... '.
-
... ....
.. ".
... _-.....
~~W~~~~
E
,.
c 22
, ',, ASiC Gate
15 I' /
, I MPU Gate
low power
10
I
1999 2001 2004 2007 2010 2013 2016
FIgure 1.1 Technology nodes and minimum feature sizes from application ITRS Roadmap:
MPU, Micro Processing Unit: ASIC, Appticalion-Spccific Integmted Circuit. Used with
pcrmission, from W. Klingenstein (2002). Technology Roadmapfor Semiconductors.
http://broadband02.ici.ro/program/klingcnstein..3d.pdf,pagelS.!OlnfineonTcchnologics AG.
2002.
complexity, microelectronic systems continue to be produced at lower costs. In Table 1.1

the costs per bit and costs per chip as well as the associated performance levels are given
as functions of the integration level. One can see that every three years the number of
bits per chip has increased by a factor of four (even faster than predicted by Moore's law)
and the cost per bit has decreased by a factor of two or more.
In the same table, the integnllion levels of logic circuits and microprocessors are
forecast. We see that, for this case, device integration is also large but will increase
slightly slower than for DRAMs. The cost of the principal clements of logic circuits-
transistors - is significantly greater, but it also tends to decrease. The forecast for the
necessary power supply presented in the table portrays a slow, but persistent, decrease.
Thus, one can expect favorable trends for the power consumption of microelectronic
systems.
The bottom of the table presents the necessary technological parameters for ultra-high
integration: minimum feature sizes, diameters of wafers, and electrical defect densities.
The large wafer size allows a greater number of devices to be fabricated on a chip. The
density of electrical defects (i.e., crystal imperfections which affect electrical properties)
is characteristic of the quality of the wafers. Table 1.1 forecasts that wafer diameters will
be continuously increased, while the defect density decreases by a factor of six per
decade; currently, they must be limited to several tens per m 2 .
After this overview of the dominant driving forces in nanoscale development, we will
mention briefly other general issues important for this field. These include improving
materials, fabrication and measurement techniques on the nalloscale, and novelties in
the operation principles of nanodevices.
Improving materials on lt1e nanoscale

In the processes of achieving minimum device sizes and ultra-high levels of integration
it is necessary to identify the limiting and critical parameters for improved performance.
In reality, these parameters depend on the integrated elements of each individual material
system. For example, for transistors two parameters of the host material are of special
importance: the ultimate electron velocity and the limiting electric field which does not
induce electric breakdown. Further improvements in the parameters can be achieved
through materials engineering.
Silicon plays the central role in electronics. However, semiconductors other than sil-
icon can be llsed. In particular, compound semiconductors constitute a general class
of semiconductors that has been used increasingly in recent decades. As examples of
forming compound semiconductors, every particular element in column 111 of the peri-
odic table of clements may be combined with every element in column V to form a
so-called III-V compound, which is semiconducting. Then, two or more discrete com-
pounds may be used to form alloys. A common example is aluminum-gallium arsenide.
A1.I"Gal_xAs, where x is the fraction of column III sites in the crystal occupied by Al
atoms, and the fraction I - x is occupied by Ga atoms. As a result, it becomes possible
not only to make discrete compounds, but also to realize a COntillllOIlS range of materials
for tailoring necessary electronic properties. As for Si technology, the growth of silicon-
germanium (Si.rGel_x) alloys facilitates the control of the properties of materials over a
considerable range of the electrical parameters. These techniques arc exploited widely
in microelectronics.
Further revolutionary modification and engineering of materials can be accomplished
by using heterostructures with nanoscale features. Helerostnlctures are structures with
two or more abrupt interfaces at the boundaries between the different semiconductor
materials. With modern material-growth techniques, it is possible to grow structures

with transition regions betwccn adjacent materials that have thicknesses of only one
or two atomic monolayers. This allows one to fabricatc multilayered semiconductor
structures with nalloscale thicknesses.
The simplest multilayered structure has a single heterojunction, i.e., a single-
beterojunction structure is made of two different materials. At the interface of such
a heterojunction. the electronic properties are changed to improve selected physical
characteristics. In particular, electrons can be confined in a thin layer near the interface.
In fact, the layers with confined electrons can be made so thin that wave-like behavior-
that is, quantum-mechanical behavior - of the electrons becomes apparent. The same
phenomena occur for diverse multilayered nanoscale structures that can be grown with
high quality.
By using nanostructures. it is possible to modify the electronic properties of a great
variety of a nanoscale devices. Indeed, we live in a tJuce·dimensional world, where a
particlc can, in principle, move in all threc directions. Quantum effects on the nanoseale
determine the properties of electrons in nanostruetures: the nanostructures can be made
in such a way that the electron motion becomes fwo-dimensional. one-dimensional. or
even zero-dimensional. These nanostructures are known as low-dimensional quantum
heterostructures and are called quantum wells, quantum wires, and quantum dots, for the
cases where the electrons are confined in one, two, and three dimensions, respectively.
Such an impressive example of manipulation of the properties of the current-carrying
particles clearly illustrates fundamentally new possibilities for electronics that become
viable at the nanoscale.
Fabrication techniques on the nanoscate

TIle progress in heterostructure technology has been made possible largely as a result of
new advances in fabrication techniques. In Table 1.2, we provide a very brief summary
of some important steps now used in the growth, characterization, and processing of
heterostructures. In the period of the 1970s and 1980s, molecular-beam epitaxy was
invented, developed, and employed to fabricate high-quality and ultra-thin layers and
superlanices. Qualitative electron-beam and X-ray microscope technologies were used to
characterize the perfectness ofstructures, including interface disorder. During this period,
lithographic and etching methods suitable for microscale and nanoscale devices were
proposed and realized. In the 1980s and later, new epitaxial techniques were developed;
these included metal-organic vapor-phase epitaxy and metal-organic 1110lecular·beam
epitaxy, among others. These innovations made possible the fabrication of layers with
atomic-level accuracy. Lithography and etching methods were improved to the point
that they can be used for nanoscalc structuring. Desirable spatial-modulation doping by
impurities has become possible, including h-doping - that is, doping of one or a few
atomic 1110nolayers.
These approaches to the production of nanostructures and nanoclectronic devices
represent "evolutionary" improvements in the growth and processing methods applied
previously in microelectronks. Advances in nanotechnology allow, in principle, the uti·
lization of methods and concepts from other areas of science and engineering. Synthetic
t
I
! ' r
nl"rl.·olc •..-'n,'o".ol.~
I II
I
Table 1.2 Advances in growth, characterizatioo. and processing of quantum

heterostructures
1970s-1980s Growth and fabrication methods

Molecular-beam epitaxy
Uhra-thin·layer fabrication
Superlanice fabrication
Characterization methods
lithogrnphic microstructuring
Qualitative electron·beam and X-ray microscopies
I99Os-2000s Growth and fabrication methods
Metal-organic vapor.p~ epitaxy
Metal--organic molecular·beam epitaxy
Fabrication to atomic-layer accurncy
~-Doping
Controlled strained layers
Fabrication methods based on chemistry and biology
Assembling inorganic nanoblocks with biomolecules
Characterization methods
lithography and etching for nanostructuring
Di~pen nanolithography
Quantitative electron-beam and X·ray microscopies
Scanning tunneling microscopy (STM)
Atomic force microscopy (AFM)
Picosecond and femtosecond spectroscopy
Terahertz time-domain spectroscopy
chemistry and even biology have much to offer for emerging nanotechnologies. Some
fundamental concepts coming from these fields can successfully be exploited for the
synthesis of nanomaterials and nanodcvices. These include chemical and biological
methods of growth of nanoscale objects - such as carbon nanotubcs and biomolecules-
surface nanopalterning, and preparing nanostructured materials with predefined, syn-
thetically programmable properties from common inorganic building blocks with the
help of DNA interconnect molecules, etc.
Improvement in characterization methods for the nanoscale

Progress in the refinement of fabrication techniques for making nanostructures depends
on the great improvements made in characterization methods. Some of these methods are
indicated in Table 1.1. In particular, composition and dopant distribution, lattice strain,
and other parameters within nanostructures must be known with atomic·scale precision.
Currently. the manipulation of a single atom (ion) in a solid is possible. New tools -
scanning tunneling microscopy and atomic· force microscopy-which portend numerous
applications in high-precision fabrication have emerged. Picosecond and femtosecond
spectroscopy have progressed substantially and they have been applied to characterize
.. ...,- ... ,,, .....-

the electronic and lattice properties of heterostructures. Finally. terahertz time-domain

spectroscopy was developed, which makes it possible to measure electric signals with
time resolution at the level of 10- 12 seconds.
New principles of device operation at the nanoscale
Fundamental questions arise when conventional principles of device operation fail as

a result of entering the nanoscalc domain. One of the effects is almost collision/ess
motion - frequently cal1ed ballistic motion- of electrons flying through short devices.
As mentioned previously, the nanoscalc domain is the "realm" of quantum physics.
Indeed, scaling down of devices and their integration above the level corresponding to
250 Mbits on a single chip makes it necessary to take into account new regimes and even
to modify the principles underlying device operation. Further device downscaling and
higher integration densities for information capacities exceeding I Obit per chip imply
the need to investigate using qualltum regimes of operation in future years. Quantum-
mechanical effects are not only important for operation of future integration circuits. but
also are already used for generation of ultra-high-frequency electromagnetic emission.
A relevant example is lhat of resonant-tunneling phenomena in nanoscale multilayered
structures, which creates a foundation for microwave emission up to 1000 OHz.
A number of such device-related quantum effects has been discovered. New physics
and new quantum effects always pass ahead of new devices exploiting these effects that
have made a substantial impact on device technology. We mention here just a few quan-
tum effects: 1970, the proposal of multilayered stnlctures; 1974, the resonant-tunneling
effect: 1978, the modulation doping effect; 1980, conduction of polymers; 1985, the
discovery of the buckyball, Coo: 1993, the discovery of single-walled carbon nanonlbes;
and 1996, nanoelectromechanical systems (NEMSs). Some of these effects will be ana-
lyzed in this book. Here is a short list of some of the novel quantum devices: 1979. the
injection quantum-well laser; 1983; the Microwave DBRTD Oscillator; 1984. the hot-
electron transistor; 1998, the quantum-wire carbon nanotube field-effect transistor; 1998,
polymer-based transistors and light-emiuing devices; 2001, sensors based on NEMSs;
2001, sub-terahertz III-V compound nanoscale field-effect transistor; and 2006, sub-
terahertz InP and SiGe bipolar transistors. There is a "delay time" between the discovery
of the effect and the device concept. but the delay time is decreasing steadily. The fol-
lowing examples show this tendency. The effect of resonant tunneling was discovered
in 1974; the device - the microwave double-barrier resonant-tunneling diode (DBRTD)
oscillator - was realized in 1983. The first quantum wires with one-dimensional electron
motion were studied in 1986; their first application in lasers occurred in 1995. In both
cases thc "delay time" was 9 years. The discovery of single-walled carbon nanotubes
was made in 1993 and in only 5 years the carbon nanotube transistor was fabricated. The
same is valid for the development ofnanoelectrolllcchanical systems and their applica-
tions for a number of sensors, etc. Thus, for contemporary electronics there is an evident
acceleration of the implementation of fundamental physical effects.
Besides quantum effects, reducing device dimensions results in a decrease in the num-
ber of electrons participating in the transfer of an electric signal. As a result, nanoscale
t ! , c! j ) II
devices may operate on the basis of single-electron transfer. Various novel single-electron
devices have been proposed and demonstrated. By reducing the sizes of quantum dots
•
to 100 A or less, it is possible to operate with single electrons at tcmperatures near or
close to room tempcrature.
The great technological advances brought about in mainstream microelectronics and
nanoelcctTOnics can be uscd for the fabrication of othcr c1asscs of nanodevices. One
such approach is based on quantum dots arranged in locally interconnected cellular-
automata-like arrays. The fundamental idea ofoperation of cellular automata is to encode
information using the charge configuration of a set of quantum dots. Importantly, in the
quantum-dot cellular-automata approach, the information is contained in the arrange-
ment of charges of the dots, rather than in the flow of the charges, i.e., electric current.
It can be said that the devices interact by direct Coulomb coupling rather than via the
current through the wires.
Another approach employs both electrical and mechanical properties of nanostruc-
tures. The new generation of devices and systems based on this approach is commonly
referred to as !/(JnoelectmmecJwlI;cal systems (NEMSs). Indeed, on the nanoscale a
strong enhancement of coupling between electronic and mechanical degrees of freedom
occurs. This electromechanical concept may be used for the development of a new class
of devices that includes nanomachines, novel sensors, and a variety of other new devices
functioning on the nanoscale. Thus, NEMSs may supplemenl the tr.lditional electronics
that works solely with electrical signals.
Nanotechnology for optoelectronics

Thus far, we have reviewed nanoscicnce and nanotechnology as applied to electronic
devices, i.e., devices in which electrical properties are exploited and which operate with
electrical input and output signals. Another class of devices is comprised of optoelec-
tronic devices, which are based on both electrical and optical properties of materials
and operate with both optical and electrical signals. An important and growing trend is
that optoelectronics complements microelectronics in many applications and systems.
First of all, optoelectronics provides means to make electronic systems compatible with
lightwave communication technologies. Furthermore, optoclectronics can be used to
accomplish the tasks of acquisition, storage, and processing of information. Advances
in optoelectronics make significant contributions to the transmission of information via
optical fibers (including communication between processing machines as well as within
them), to the high-capacity mass storage of information on laser disks, and to a number
of other specific applications. Clearly, optoelectronic devices have a huge number of
diverse applications.
The principal components ofoptoelectronic systems are light sources, sensitive optical
detectors, and properly designed light waveguides, for example, optical fibers. These
devices and passive optical clements are fabricated with optically active semiconductor
materials. Semiconductor nanostructures and, in particular, quantum heterostructures
provide new means to enhance many optical and electro-optical effects. For example,
both of the most widely used light sources - light-emitting diodes and laser diodes -
t c !, II
Illay be improved greatly when nanostructures such as quantum wells, quantum wires,
and quantum dots are exploited as active optical elements.
As for the previously studjed case of microelectronics, the trends in optoelectronics
involve scaling down the sizes of these devices as well as achieving high levels of
integration in systems such as arrays ofJight diodes, laser arrays, and integrated systems
with other electronic elements on the same chip. Optoelectronics benefits substantially
through the use of nanotechnology and becomes competitive with its microelectronic
counterpart.
In conclusion, the current and projected trends in electronics lead to the use of nanos-
truclures and to the reliance on novel quantum effects as an avenue for realizing further
progress. These recent and diverse tfends in semiconductor and device technologies as
well as in novel device concepts are driving the establishment of a new subdiscipline
of electronics based on nanostructures, i.e., nanoelectronics. This subdiscipline and it's
foundations are studied in this book.
More general information on nanoscience, nanotechnology, and nanostructures, and
their potential, may be fowld in the following reviews:
R. Feynman, 'There's plenty of room at the bOllom," American Physical Society

Meeting, Pasadena, CA, 29 December 1959; originally published in Caltech's
Engineering and Science Magazine, February 1960; reprinted as R. P. Feynman,
"Infinitesimal machinery," Microelectromeclwnical Systems, 2, I (1993); (see, for
example, www.zyvex.com/nanotech/feynman.html).
National Nanotechnology Initiative: The Initiative a"d Its Implementation Pia",
ational Science and Technology Council, Committee 011 Technology, Washington
DC, 2000 (see, for example, w\vw.nano.gov).
H. Kroemer, "Quasielectric fields and band offsets: teaching electrons new tricks,"
Rev. Mod. Phys., 73, 783 (2001).
The International Technology Roodmap/or Semiconductors (Semiconductor Industry
Association, an Jose, CA, 2002 - update).
,,
"','7i~olc •.._ 'n'~_1nl~""~
II
I
2 Particles and waves
2.1 Introduction
The evolution of microelectronics toward reduced device sizes has proceeded to a degree
that renders conventional models, approaches, and tJlcories inapplicable. Indeed, for
objects with sizes of 100 nanometers or less it is frequently the case that the length scales
associated with fundamental physical processes arc comparable to the geometrical size
of the device; also, fundamental time scales associated with physical processes are of
the order of the time parameters for nanodevice operation. Therefore, on the nanoscale
the theories and models underlying modern nanoelectronics become more complicated,
and rely morc and more on basic science.
Generally, in the nanoworld the fundamental laws of physics that govern particles and
material fields differ from those that apply to familiar macroscopic phenomena such as
the mOl ion of a baseball or a train. Instead of classical mechanjcs, that works so well
for macroscopic phenomcna, the motion of particles in the nanoworld is determined
by the so-called wave mechanics or quantum mechanics. An underlying principle of
central importance for nanophysics is the fundamental concept that all maller, including
electrons, nuclei, and electromagnetic fields, behaves as both waves and particles, that
is, wave-particle duality is a basic characteristic of all maUer.
At first glance, wave properties and particle properties for the same physical object
are hardly compatible. To understand wave-particle duality, we will briefly review, in
the following two subsections, the basic properties of particles and waves known from
classical physics.
2.2 Classical particles
A particle can be characterized by the momentum vector p and the kinetic encrgy E that
depends on the momentum. Here and throughout this book we will usc the bold fonts for
vectors, i.e., p is the vector and p = ]pl is its absolute value. The change of momentum
with time is defined by Newton's second law:
dp
dr =r. (2.1)
where t is time and f is the vector of an external force. From Eq. (2.1) it follows that, if
the force is absent, then dp/dt = 0, i.e.. p = constant. This is the so-called momentllm
t I c!!' ,j I
conservation law valid for a mechanical system in the absence of external forces. For
simplicity in classicalmcchanics, we assume that any particle we consider has a very
small size in comparison with the space where the particle is located. We refer to such a
particle as a point particle.
111e coordinate vector, r, ofa point particle and the particle velocity, v, arc related by
the well-understood relation
dr
- = v. (2.2)
dt
To obtain the relationship among the velocity, v, the momentum, p, and the energy of a
particle, E, one has to calculate the power associated with the force f (the work of the
force f on the particle per unit lime). So, we have to multiply the left~ and right-hand
sides ofEq. (2.1) by v, resulting in vdp/dr = fv. (Remember that the scalar product of
two vectors a and b is defined as ab = a,.b.y + a.loby + lI:;b:;.) The right-hand side of the
tr.:msformed version of Eq. (2.1), fv, is equal to the rate of the energy change dE/dt
(dE/dt = fv), and we obtain the relation
dE dp
dI = v dr' (2.3)
Using the chain rule of function differentiation (dJ(x)/dt) = (dJ /d\"")(dx /dt), we deter-
mine how the velocity of the particle is related to the momentum and energy:
dE
v-- (2.4)
- dp'
Here, the derivative with respect to the vector p also gives the vector v with components
dE dE dE
lJx = d' vy = d' and u, - --
- dp,'
p" py
Let us consider an important case of a particle moving in a potential field. The force
is defined as the derivative of a potential V(r) with respect to the particle coordinate:
r = -d V /dr. Note that, forthe vector operntor d/dr '" {d/dx. d/dy. d/dz}, ooe often
uses another notation: d/dr:::: V, so that d V /dr is the so-called gradiellf of function
V(r), V V(r). On multiplying the left- and right-hand sides of Eq. (2.1) by v, using the
definition of Eq. (2.2) and the chain rule, we find
dp dV dr d
v - +- - =- (E + V(r)) = O.
d1 dr dt dt
The value of the kinetic energy plus the potential energy,
1i = E + V(r). (2.5)
represents the total energy of the particle, H. The above calculations tell us that the
total energy of a particle in a potential field does 110t change during its motion. So, we
have demonstrated the law ojenergy collsen1ation, d11/dt = O. When 1t is considered
as a function of two variables p and r, it is called the Hamiltonian fUl1ction. or simply
the Hamiltonian. Remarkably. the partial derivatives of 1-l give us both fundamental
equatinns (2.1) and (2.4): dp/dt = r = -a1i/ar and dr/dt = v = a1i/ap. As we shall
. , , "1
2.3 Classical waves 13
sec, the Hamiltonian of classical physics also plays an important role in the formulation
of quantum mechanics.
A point particle moving in free space may be characterized by a mass m and by the
kinetic energy:
p'
E= - . (2.6)
2m
The latter dependence is frequently referred 10 as the energy dispersion. Here £ is an
isotropic function ofp. From the definition ofk.inetic energy given by Eq. (2.6) we find
that
p
v =-. (2.7)
/II
So, the velocity and the momentum are collinear vectors. Then, Nc\vton's second law,
Eq. (2.1), can be rewritten in its usual form:
d2 r
m- =[ (2.8)
dt Z
Now, Eq. (2.5) takes the form that we will use oftcn in this book:
p'
'It = -2m + VCr). (2.9)
One of the important results following from classical mechanics is that, ifwc know the
particle position ro and its momentum Po (or velocity yo) at an initial moment 10, from
Eqs. (2.1 )-(2.4), we can find the position and the momentum (velocity) of the particle
at any given moment of time' for any given for VCr).
Equations (2.1 )-{2.8) are the equations of classical mechanics. All of the variables,
such as r, p, E, and v, are continuous variables. Importantly, Ipl can have any value,
including zero, i.e., p = 0 and E = 0 are allowed.
For a particle, sayan electron, moving inside of a crystal (a metal, a dielectric, a
semiconductor, etc.), the interaction of this particle with the crystal generally makes the
relationship between E and p - the dispersion relation - more complicated. In particular,
E may be an anisotropic function, and the velocity and momentum may be noncollinear
vectors. Examples of such energy dependences will be given in the problems for this
chapter.
2.3 Classical waves
We are all familiar with a lot of examples of waves and wave processes. These include
sound waves in air, sea waves, and elastic waves in solids, electromagnetic waves, and
gravitational waves. Generally, in classical physics wave motion arises in extended con-
tinuous media with an interaction between the nearest elements of the medium. Such
an interaction gives rise to the transfer of a distortion (an excitation) from one ele-
ment to another and to a propagation of this distortion through the medium. Despite the
11-2 11-1 11 11+1 11+2
z
•
ZIl_2 = (11 - 2)(1 Z,,_l = (n - 1)(1 Zn = na Zn+1 = (11 + I)a 2',1+2 = (11+ 2)a
FIgure 2.1 A linear chain of identical atoms of mass AI: II" arc displacements of atoms from .heir
equilibrium positions and the restoring force acting on the 11th atom is I" = -f3(II" - 11,,+1)-
fJ(ll n -11,,_1). Note: displacements II" are not shown to scale.
differences in the particular nature of waves, wave motion has much in common for
different media. We introduce wave properties by analyzing the following simple model.
We will construct a model of a one-dimensional medium, clement's of which are
represented by "atoms" connected by massless springs. Vibrations in such a linear atomic
chain are governed by the laws of classical mechanics. The chain is supposed to be
infinitely long. Let the equilibrium distance between atoms be fl. Thus, the equilibrium
position of the 11th alom is Zn = IW, and the displacement of this atom from its position
is denoted by Un. Figure 2.1 depicts such a linear chain of identical atoms of mass M.
The springs represent interatomic forces, i.e., interaction between nearest elements of
the medium. If the displacements of atoms from their equilibrium positions are not too
large, the restoring forces in the chain obey Hooke's law,
f=-fJu. (2.10)
where /I is a change of the spring length, fJ is the spring constant, and I is the force
exerted by the spring. Now, we can apply Eq. (2.10) for the total force, In' acting on the
11th atom coupled with its two nearest neighbors by two springs as
In = -fJ(u n -Un+l) - fJ(u n - lin-I). (2.11)
Hence, the Newton equation of motion, Eq. (2.8), for the 11th atom is
d2u n
M , = -fJ(2u n -/1,,-1 - Un+I). (2.12)
dr-
This set of linear differential equations, Eq. (2.12), in principle, describes wave·like
processes. However, we will make a further simplification and modify this discrete set
of eqlffitions to obtain one equation describing a continuous medium. Such a continuous
medium with elastic forces between its elements is, obviously, a string. To make this
transformation to the continuous case, we shall consider the discrete coordinate to be
continuous, Zn ~ Z, and replace the finite difference in Eq. (2.12) with a derivati ve:
Un -Un_1
""--="'- -> -
aLi
and
(2/1 n -lin_1 - I1 n +l)
~
a
--
2
u
(2.13)
(I az 02 azl .
Then, we obtain the equation describing the displacement, II, ofa string:
a2 /1 a2 /1
(2.14)
pJD ---, - A lD ---, = O.
8t- (Jz-
where PID = Mfa is the linear density oftile string, and A I D = fJa is the so-called clastic
modulus of the sIring. The infinite sel of ordinlllY differential equations, Eq. (2.12), is
replaced by a single partial differential equation. Since we started our derivation for a
mechanical system, for which the energy (per atom) can be defined, it is instructive to
find a similar characteristic for the continuous medium described by Eq. (2.14). Assume.
for a moment, that the spring under consideration is of finite large length, L. Then, let
us multiply Eq. (2.14) by ali /81 and integrate it over the length L:
lo
' (d:
alII -au
Pm -
ar l at -
a 211 au)
oz2 -at = o.
A to -
By integrating the second term by parts by using the standard relationship J w dv =

wv - J v dw. we find the following identity for a unit length of the string:
:,~ t dz[PtC';j + A~DC~)']

_ AID
L
[(a,~
a~
au)
a, _(a,~ all) ]
,~, a, a~ ,~O
= o.
To draw further conclusions, we define an average of a quantity, A, over a piece of the

string, Z, as A = (1/ Z) ~l) Adz. For a long string, the average should not depend on
the length, Z, of the piece of string. Now, as L ~ 00, the latter identity will be satisfied
if the value
PIO
"ID= - (all)'
2 a,
- +--
2
(all)'
--
az
AID (2.15)
is independent of time, i.e., all 10/81 = 0, and the value II 10 is conserved. In fact, the
first term on the right.hand side of Eq. (2.15) is, obviously, the density of kinetic energy
over the length of the string, while the second term is the clastic (potential) energy. Thus,
II 10 has the meaning of the energy density for our one-dimensional continuous medium.
As expected, the ener!,')' density is conserved, if external forces are absent.
Our "one-dimensional" analysis, which assumes that the atoms can move only along
a single direction I, can be generalized to a three-dimensional elastic medium; see
Problem 4 of this chapter. Now, the displacement becomes a three-dimensional vector,
u, and, instead of Eq. (2.14), we obtain
a'aru (a'u
p -1- A -
ilx
+a'
-
ay
u2
+a'u)
az- =0 2 l
(2.16)
with ... = {x. y. z} being the vector coordinate and p being the "three-dimensional" den-
sity, i.e., the mass of a unit volume of the medium; A is the elastic modulus of the
medium. Since it was introduced as a single elastic modulus, independent of direction,
Eq. (2.16) is valid for an isotropic medium. The energy density of a three-dimensional
clastic medium is
, ,
(au)-
It = P
2 at
+ (au)
2 ar
A (2.17)
Here, the bar denotes an average over a small volume l:1 V(A = (II f1 V) ~lJ..l') A d V),
and
au Bu x alt y all; .
- '" - + - + - = dlv u.
ar ax ay az
Finally, we can rewrite Eq. (2.16) in the standard form of the wave equation:
alu ,(8
<.J(2 - s- ux1
2
U a2 u
+ ayl + azl
alu)
= 0, (2.18)
where we introduce a new parameter s, the meaning of which will be clarified below. For
Ollr model
.<=jll/p. (2.19)
Although we derived Eq. (2.18) for a particular model of tile elastic medium, lhe equation
can be applied to describe a wide class of physical vector fields as exemplified by
u(x, y. z) - the displacement field associated with a wave in an elastic medium. If the
characteristic of a wave field is a scalar value, say w, in Eq. (2.18), we should simply
substitute u ~ w to describe the case of a scalar physical field.
Now we will analyze solutions to Eq. (2.18) for some cases in which the solutions
are particularly simple. In many cases, such solutions are associated with wave·like
processes. We may look for solutions of the form
U(1. r) = A cos(qr - WI) + B sin(qr - WI). (2.20)

where A and Bare arbitrnry vectors, wand q are unknown parameters; w is known as
the allglllarjrequency of the wave and q is called the l1'avevector. By substituting such a
form for U(I . r) into Eq. (2.18), we easily fi nd that Eq. (2.20) is a solution of Eq. (2.18),
if w 2 sq2. The relationship between wand q
= =
Iql is called the dispersion relatioll:
W = .<Iql. (2.21 )
Importantly, there is no limitation to the wavevector q: a solution can be found for any
q. This is valid o"Iy for infinitely extended media. Jor which the wavel1ector can be a
"continuous" vector.
Because the two terms in Eq. (2.20) behave similarly, we can discuss basic properties
of these solutions based on the example of "sinusoidal" waves:
U(I. r) = Bsin(qr - WI). (2.22)

The argument of the sine function is the phase of the wave, ¢ = qr - wt, and B is the
amplitude of the wave. Let the coordinate r be given, then we obtain a function that
oscillates in time with an angular velocity w. The frequency defines the rale of change
of the phase with time t (radians per unit time). The period of time associated with a
single oscillation is
T = 2rr/w.
Accordingly, T is known as the period. If the lime t is fixed, Eq. (2.22) represents a
function that oscillates as the coordinate changes. These oscillations arc characterized
by the wavevector. q (or wavenumber, q). The wavevector defines the rate at which the
11 11
, ,
-:, A=2Jr Iq :-
,
8 , ----------- , 8
,,
,,,
,,
-J. =f ,
\--r-/'--t-\--i-;;1--t-\--r-/4-+z
o --1. J.
4
-8 -8
(a), ~ 0 (b)'~TI4
Figure 2.2 A propagating W'dve II = B sin(qz - WI). Al time t = 0 the wave It = B sin(qz - wt)
al the point z = -J./4 has displacement II = B sin( -Tr /2) = -B. At time t = T /4 we have
lJJ1 = (2Jr I T)(T j4) = ;r /2; the same displacement will occur at the point z = 0: /I =
Bsin(qz -wi) = Bsin(-rr/2}= -B.
phase changes with the coordinate (radians perunillength). One can introduce the spatial
period to represent the distance for which u undergoes an oscillation of one cycle. It is
called the wavelength,
A = 27r/q. (2.23)
The relationship between the time period and the spatial period is defined by Eqs. (2.21)
and (2.23):
27r
A = - s = Ts. (2.24)
w
The waveforms of Eqs. (2.20) and (2.22) are often referred to as traveling plane
waves. Indeed, their phase changes in a single direction along the wavevector q and
surfaces of constant phase are planes perpendicular to q. Thus, in the case of a plane
wave in a three-dimensional medium, the wave parameters do not depend on the two
coordinates perpendicular to q. A traveling wave is illustrated in Fig. 2.2, where this
wave is shown at two different moments of time. Now, we can clarify the meaning of
the parameter s. Let the wavcvector be directed along the z direction. Then, the wave
phase is ¢ = qz - wI = q(z - (w/q)/]. So we can conclude that a given magnihldc of
the phase moves with the velocity w/q = s, i.e., the parameter s in the wave equation
(2.18) is the phase velocity of the traveling wave. In the waveform of Eq. (2.22), the
constant B is the wave amplitude. The average of "2(/. r) can be expressed in terms of
Bas u 2 = 8'1/2, where we use the formulae [sin(qr - wt)]2 = [cos(qr - wt)]2 = !.
According to Eq. (2.15), the amplitude relates to the wave energy density_ Indeed, we
note that
p au A au
() ()"
-2 -a, ' + -2 -ar =
8-,
-[pw-
2
,") ,
[cos(qr - wt)]- + '/"" [cos(qr - wt)]-]
= ~'(p~' +q:,,) = ~'(p~' +p;') = ~'pw'

In obtaining this result, we have taken into account that q2 = w 2 /s 2 and s = ,j Alp.
TIlUS, we obtain" = pw 2 8 2/2. Importantly, the energy density for traveling plane waves
is independent oJllte coordinate and it is proportional to CJ;2. If w = 0, there is no wave
and there is no energy associated with it. Note that, for the waveforms of Eqs. (2.20) and
(2.22), averaging over a small volume of the medium is equivalent to averaging over the
period of oscillations.
For trJvcling waves, one often introduces also the wave intensity. which is the density
of the energy flux. This energy flux represents the energy transferred by the wave per unit
lime through a unit cross· section perpendicular to q. The intensity is a vector directed
along q with the absolute value
(2.25)
Besides the waveform of Eq. (2.20), a traveling wave can be presented in a complex
form,
u = Aei{qr-wl). (2.26)
where the amplitude A is generally a complex quantity. Indeed, Eq. (2.26) is a solution to
the wave equation (2.18). For some cases, it is convenient to operate with such a complex
waveform. However, one should remember that true physical quantities always have real
values.
Now we consider the important wave phenomenon known as wave il1lel!erence. Sup-
pose that two sinusoidal waves of the same frequency propagate from different sources
through the medium. The sources of the waves arc generally at different locations, so
the waves reach a point of observation r, in general, with different phase shifts 'P1(r) and
I',(r):
0,(1. r) = B, sin(wl + I',(r». 0,(1. r) = B, sin(wl + I',(r)). (2.27)
The resulting wave field is u(r, r) = u dr, r) + 1I2(r. r). In experiments, instead of
the wave amplitude, it is the intensity of the wave that is measured in many cases.
- -
From Eq. (2,15) the intensity is proportional to u 2• Straightforward calculation of u 2
gives us
,.
0- = 21("
H, + Hi + 2B, B, cos [I',(r) - I',(r)1) . (2.28)
In deriving Eq. (2.28), we have used the identity

.. 1 I
Sin X Sll1Y = 2 cos(x - y) - 2 cos(x + y),
I.e"
and the fact that the average of COS(2Wl + 'PI (r) + 'P2(r» is zero. Thus, the intensity
of the resulting wave consists of three contributions: the term related to the wave
Light
Agure 2.3 A double-slit experiment: 51 and 51 form the double slit.
coming from the first source, the term related to the wave from the second source,
and an additional tenn. This third contribution describes the effect of wave inter-
ference. It depends critically on the wave phase difference. The interference contri-
bution can be positive (col/struclive interference at -Jr/2 < (~I - rpz) < rr/2 when
cos (fPl - ~2) > 0), or negative (destructive interference at Jr/2 < (~l - rpz) < 3rr /2
when COS(~l - rpz) < 0). Lmportantly, the interference effect can be observed only
for waves with the same frequency (otherwise the averaging leads 10 zero interfer-
ence contribution). The waves of Eq. (2.27) with time-independent phase shifts are
also known as coherell1 waves. If waves are characterized by a phase shift (/PI -/Pz),
which jitters randomly in time, the waves are incolterellt and no interference effect
occurs.
The simplest example illustrating the interference effect is the double-slit optical
experiment Let two slits be illuminated by a light wave from a single source. These two
slits become t"vo sources of coherent waves. The superposition of these waves generates
an interference pattern of fringes, as shown in Fig. 2.3.
The wave analyzed above travels along the vector q. Using the wavevector -q in
Eq. (2.22) at the same frequency, we obtain another wave traveling in the opposite
direction. According to Eq. (2.20), a combination of these waves is also a solution to
Eg. (2.18):
o(t. r) = B+ sin(qr - wt) + B_ sin(qr + wI).

(a) II = I (b) II = 2 (c) 11=3

FIgure 2.4 Quantization of oscillations in the form of standing waves. The gray area is a
standing.wave pattern. Solid lines show amplitudes of oscillations at an instantaneous moment
ohime. Only waves with integer half-wavelengths exist L = nA/2. Note: for demonstration
purposes. the amplitudes are shown not to scale.
where again B+ and B_ are arbitrary constant vectors. The two waves can also be
interpreted as incident and riif!ected waves.
An important case is that of a wave propagating between 1\\'0 reflecting walls
placed at z = 0 and z = L. The waves depend on a single coordinate z: U(I. z) =
B+ sin(qz - WI) + B_ sin(qz + WI). If the walls are rigid there are no displacements
at the walls, so we should use the boundary conditions for the waves at z = 0 and
z = L: u(z = 0) = u(z = L) = O. One of the many physical embodiments of this sit-
uation are represented by waves in a continuum elastic medium bounded by rigid
boundaries at z = 0 and z = L; another is given by a string attached to two fixed
boundaries at z = 0 and z = L. From the boundary condition at z = 0, we find
B+ - B_ = O. thus U(I. z)= B+[sin(qz - WI) + sin(qz + WI)] =
2B+ sin(q=) COS(wI).
(Note: sinx + siny = 2 sin[lCx + y)]cosr!(x - y)]). The boundary condition at z = L
can be satisfied if, and only if, sin(q L) = O. The latter requires the so-called "quantiza-
tion" of the wavevector q, q L = 7f", or
1T
q = {fit = L 11. 11=1.2.3..... (2.29)
where qll is called the wavenumber. That is, there exist only waves of a special form,
u = B sin(qnz) COS(WIt/). (2.30)
at discrete values of the wavevector qn and frequency W n = sqn. This important class of
waves is known as sta"ding waves. From Eqs. (2.23) and (2.30), we have the relation
L = (Aj2)n. For standing waves, a strictly integer number of half-waves may be put
between the reflecting walls. Figure 2.4 illustrates standing waves of various wavelengths
for an arbitrary instant ohime. The longest possible wave (11 = I) corresponds to A = 2L
when a half-wave fits between the reflecting walls; see Fig. 2.4(a). To calculate the
energy density of the standing waves, we can lise Eq. (2.15). Then, the energy density is
"II = pB2w~/4. It is independent of coordinate, as found for the traveling plane wave.
Obviously, the energy flux is now zero. It is important to stress again the difference
between traveling waves (with arbitrary and continuous values of q including q -+ 0,
hence >.. ~ 00, in infinite media) and standing waves with quantized (or discrete) values
of qll with the minimum wavenumber ql = 1r / L.
The waveform given by Eq. (2.20) is actually a simple specific example of a more
general class of wave fields. Indeed, due to the linear character of the wave equation,
a sum of an arbitrary number of partial solutions is also a solution of the equation.
This property of \wves satisfying linear wave cquations is known as the slIpeipositio"
pri"ciple. According to this principle, we may write the solution to Eq. (2.18) in the
general (complex) form as
U(/. r) = L Aqei(qr-wql). (2.31 )

q
Here, the summation is taken over the wavevectors q. In addition, w q relates to q through
the dispersion relation (2.21). The amplitudes of waves contributing to U depend, in
general, on the waveveclors. The superposition principle is the basis for many important
phenomena, including interference, formation of standing waves, and diffraction.
Electromagnetic waves in free space

One of the most important examples of waves is that of electromagnetic waves, i.e.,
oscillating electromagnetic fields. These fields are responsible for the most basic prop·
erties of matter from the nano\Vorid to the scale of the universe, and they are exploited
for a nwnber of technologies critically important for modern society. The fundamcntals
and applications of electromagnetic fields constitute a separate and extremely important
field of science. In this book, we will deal with electromagnetic fields only briefly in
order to illustrate the general character of the wave equation (2.18).
Electromagnetic waves are joint electric and magnetic fields that oscillate in both space
and time. In the simplest homogeneous case, both the electric field F and the magnetic
field H arc governed by the wave equation in the form (2.18), where one should replace
u -+ For Hand S -+ C, where c is the velocity of light in free space and w = qc. The
wave equation, for example. for the electric field reads as
8'F2 _ c' (8 2 F2 + 8'F + 8'F) = O. (2.32)

ot ax oy2 oz2
Now, one can write the electric field in the form of a plane wave similar to Eq. (2.22):
F(r. I) = bF" sin(qr - wI).
where Fo = bFo• Fo is the amplitude of the electric field, and b is the vector of the
polarization of the wave which denotes the direction ofFer. r). The parameters q and w
have the same meaning as above: the wavevector and the angular frequency of the wave.
Alternatively, it is possible to usc a complex form of the plane wave,
F(r, t) = _ibFoeililr-wt), (2.33)
but only the real part of this formula has physical meaning. The same kind of equation
may be written for the magnetic field H.
Direction of
propag:l1ioll
FIgure 2.5 A plOI of electric, F', and magnetic. 1-1, fields as functions of z allime 1 = O.
In free space, the vector F is always perpendicular to q, so that, if q is fixed, the

electric field has, in general, two possible polarizations a/the electromagnetic W(lve thai
correspond to the two orthogonal directions perpendicular to q. The magnetic field vector
H is, in turn, perpendicular to both q and F. Propagation of the electric and magnetic
fields is illustrated by Fig. 2.5.
Equations of the form of Eq. (2.32) are valid for a homogeneous or nearly homo-
geneous dielectric medium, with only the replacement c ~ cj.li, where f is the dielec-
tric constanl of the medium:
a'F c' (a'F a'F a'F)

;)/2 - -;- aX2 + a 2 + azl = O. (2.34)
y
Since € > I, the velocity of light, cj Ji. in a dielectric medium is less than thaI in free
space. The frequency, w, and wavevector. q, are related through the dispersion relation
c
w= "r.lql.
The energy of the wave can be characterized by the density of the electromagnetic
energy, which is defined as
(2.35)
whcre F2(f) represents the time average of F 2(t) and EO is the permittivity of free space:
fO = 8.85 I x 10- 12 F m- I (I F = I C V-I) . We can define the intensity of the wave as
the energy flux through unit area perpendicular to the wavevector <j:
c ,
I = 2 EOE Fa· (2.36)
The above description is associated with the classical descriptio" of the electromagnetic
fields.
In closing this section on classical waves, we summarize that wave motion occurs
in extended media and waves themselves are always dclocalized physical entities that
generally extend over large distances.
2.4 Wave-particle duality 23
2.4 Wave-particle duality
In this section we will study how any physical object may behave simultaneously as
a particle and as a wave. To understand this so-called wave~parric1e duality, we use a
"two-way road" from waves 10 particles and from particles to waves.
From waves to particles
We start with a well-known optical example, which will help us 10 overcome the gap
between classical waves extending over space and discrete particles having certain coor-
dinates. Visible light behaves as an electromagnetic wave exhibiting interference, diffrac-
tion, etc. On the olher hand, we often sec light as a light ray, which can propagate, be
reflected, and be deflected. The trajectory of such a light ray is very similar to a par-
ticle trajectory. The class of optical effects for which light can be described in terms
of rays constitutes the discipline of geometrical optics. The transition from wave optics
to geometrical optics is justified if the wavelength of light A is much smaller than the
characteristic scales of the problem being considered: A « L, where L can be a distance
of light propagation, the size of an illuminated object, a scale of inhomogeneity of the
medium, etc. Consider the last case with a smooth inhomogeneity, which appears as a
"smooth" coordinate dependence of the dielectric constant €(r) in Eq. (2.34). We will use
a waveform similar to Eq. (2.26): F = Ae~. For monochromatic light (light ofa single
frequency) we can write ¢ = ¢o - WI. Here, 4>0 depends only on r. We can introduce the
wavevcctor again as q = 8¢J/8r = 8t/Jo/fJr. In an inhomogeneous medium the wavcvec·
tor and the wave amplitude depend on the coordinate vector r: q = q(r); A = A(r). On
substituting the chosen waveform, F = Aei.P, into Eq. (2.34) we obtain
, i.«r)
(a'A
aA
-w-Ae - -C'- --")c +21--<
ax-
i. .ax a<Po..
-A -
ax
(a<Po)' e'" + ... ) =0.
ax (2.37)
where, for simplicity, we have written only the derivatives with respect to the x-
coordinate. Within the geometrical-optics approximation, the phase ¢J(t. r) is large:
I¢JI » I. Moreover, both contributions to the phase are large too: WI, lcPol » I. Since the
phase is large, the leading terms in our calculations, namely the very first term (propor.
tiona I to (o¢J/at)2 = w 2) and the third term in the brackets (proportional to (8¢o/ax)2),
are quadratic with respect to the phase derivatives. Keeping these leading terms and
cancelling out common multipliers, we obtain an equation for ¢o:
( a<Po)'
ar
= w' «r).
c2
(2.38)
Ifwe find a solution ofEq. (2.38), we can define the so-called WOlle surfaces on which cPo
is constant: ¢OCr) = constant. For a given point r, the direction of the wave is detemlined
by the wavevector q = d¢o/dr; the wave direction is perpendicular to the wave surface.
After we have found the wave surfaces and calculated q(r), we may construct the ray
trajectories, as presented in Fig. 2.6. In the simplest case of a homogeneous medium,
ray
a -....
b_ - - -....- - 1 - - ray
c _....
...............
I
ray
FIgure 2.6 A sketch of the construction afray trajectories by using the wuvcfronts. Rays a. b. and
c arc pcrpendicular to the wavcfronts 1,2. and 3.
where f: does not depend on the coordinate, we obtain a trivial result: q = constant, i.e.,
the ray propagates along a straight path.
Thus, if the wavelength is small in comparison with the characteristic scales of the
system being considered, wave motion can be described in terms of ray trajectories. Up
10 this point, our discussion has been based totally on classical physics.
Thus far, we have not discussed the wave amplitude (or the wave intensity). The wave
amplitudes enter the equations only as parameters that may be defined if the parameters
characterizing the lighl sources are known. However, the wave amplitude is directly
related to the kind of interpretation we arc using: a classical interpretation or a quantum
interpretation. Indeed, it has been established that the classical-physics approach is valid
only for large intensities of waves. If the intensity of a wave is small, other laws of
quantum physics define the wave field.
Let us consider briefly basic quantum concepts related to wave fields. According to
Max Planck and Albert Einstein, electromagnetic waves interacting with matter can be
emiued and absorbed in discrete energy portiolls (quanta) - photons. The energy of a
photon, E, is proportional to the frequency of the wave:
£ = tiw. (2.39)
where tl = 1.06 x 10- 34 J s (Joule-second) is the so-called reduced Planck COllstallt;

h = 2rr 1'1 is called simply Pltlllck sconstallt. Each photon, like a particle, hasa momentum
p= t,q. (2.40)
where q is the wavevector. (Equation (2.40) is formulated for photons in free space.)
Apparently, the direction of Ihe momentum vector p coincides with the direction of
propagation of the wave. For each of the two possible polarizations of the radiation, one
introduces appropriate characteristics of photons: each polarization of light, b, is related
to a certain photon. Thus, a photon can be labeled by two indices, (q. bl.
Table 2.1 Acomparison between classical and quantum quantities
Classical quantity Corresponding quantum quantity
Density of optical energy, $V Number of photons, Nq•b = IVV/(ftw)

Optical intensity, I(r) Photon-Aux density, I(r)/(nw)
Total optical pmvcr. P Photon Aux, P/(Tlw) = Nq,bC
Equation (2.39), which represents the energy of a quantum of light, may be readily
generalized for electromagnetic waves in a dielectric medium with dielectric constant~:
fie
t,w = .,(iq. (2.41)
It is very important that different quanta of an electromagnetic field do not interact with
each other, as is reflccted by the linear character of the field in free space. An interaction
between these modes is possible only in special media. Such media are called nonlinear
optical media.
By introducing quanta, i.e.) particles of electromagnetic field, we may understand
Eq. (2.38) and Fig. 2.6 in a new way. Let us multiply this equation by the factor tl 2/(211I),
where m is an inessential parameter. Then, we may write
p2 1'I2lLi€(r)
-- =0
2m 2mc2
upon taking into account the relations q = d4'o/dr and p = 1'Iq. On defining a "potential"
energy as VCr) = -ft 2w 2€(r)/(2mc?), we come to the equation 1t :::: p2 1(2m) + VCr) =
0, which is a direct analogue of the energy conservation law for a particle (compare this
with Eq. (2.9». That is, Eq. (2.38) describes trajectories of quanta of the wave field.
In quantum theory, instead of the wave intensity one introduces the "umber ofquanta
(or phOIOIlS), Nq . b . If each of the quanta transfers the energy t,W, then the intensity,
defined as the energy flux density, is I = fiwcNq.bl V. where V is volume and Nq.bl V is
the densityofquanta with wavcvectorq. Because the quantum pichlre has to coincide with
the classical picture for a large number of photons, one can match the latter relationship
and Eq. (2.36) for the classical intensity. From this comparison, it is possible to find the
relationship between the classical amplitude of the electric field, Fo, and the number of
quanta:
2tJwNq.b
Fo = (2.42)
~O~V
Our discussion of the relationship between electromagnetic waves and photons pro-
vides an example of the wave-particle duality that is ubiquitous in quantum physics. A
comparison of some characteristics of electromagnetic fields in the classical and quan-
tum interpretations is given in Table 2.1. For a wave of small intensity (small number
of photons, Nq.b ) the quantum description is more suitable, whereas for a substantial
intensity (large N q . b ) a classical wave interpretation may be used.
From particles to waves

Consider a particle, sayan electron. On the basis of classicalmcchanics one can attempt
to characterize it by a mass, m. and a vector representing the momentum, p. In classical
physics we can know with certainty that a particle is at a certain position in space, .... From
a quantum-mechanical point of view a particle is characterized by the W(lVejilllctiol1, 1fJ(r).
Frequently, quantum mechanics is referred to as "wave mechanics." The wave function
is, in general, distributed in space. The main property of the wavefunction is that the
value 11fr(r)12 dr gives the probability of finding a particle inside of a small volume
dr around point f. Thus, the wavefunction, if!, may be interpreted as the probability
amplilude corresponding 10 Ihe probability density for finding a particle at a particular
point of space, r. That is, there is a direct analogue between 1l/F1 2 and the square of
the electromagnetic wave magnitude lFol 2, which determines the photon density, as
discussed previously.
Probabilistic behavior is one of the key features of quantum mechanics; thus, a word
of explanation is necessary 10 define what is meant by "probability" in this context.
To understand probability in a quantum-mechanical context, it is convenient to have in
mind the following situation. Imagine an enscmble of similarly prepared systems. By
"similarly prepared" we mcan identical systems as far as any physical measurement is
concerned. Now, if a measurement is made on one of the systems to determine whether
a particle is in a particular volume element, the result will be definite: either a part'icle is
there. or it is not. When the same measurements are made on a large number of similarly
prepared systems, the number of times a particlc is found in thc fixed volumc is taken
as a measure of the probability of finding a particle in the elementary volume.
For the simplest case ofa particle in free space, the wave function has the form ofa plane
wave, as introduced previously:
(2.43)
where = kr - Qr is the phase, k is the wavevector of the particle, A is the amplitude
of the wave, and Q is the angular frequency associated with the energy of the particle.
TIle wavevector, or more precisely its magnitude k = Ikl, is relatcd to thc wavelength of
a particle, J,.:
2"
-
).-
- k . (2.44)
According to the dc Broglie relationship, the momentum of a particle is related to a

wavelength associated with the particle through the equation
, -21ftl
A -_ • (2.45)
p
From Eqs. (2.44) and (2.45) we obtain a relation between the wavenumber and the
momentum of a particle: p = tlk, or, in the vector form,
p = Ilk. (2.46)
This relationship coincides exactly with Eq. (2.40) for quantization of electromagnetic
waves. This simplest case demonstrates the way in which one may attribute to a particle
both particle-like and wave-like properties.
Within the plane-wave description of a free particle, it follows from Eq. (2.43) that
there is an equal probability of finding a particle in any point of space:
, ,
III'(L 1)1- = II'(L/)II"(L/) = IAI- = constant.
where \.II·(r. f) is the complex conjugate of \.II(r. 1).

This result appears 10 be in contradiction with the classical description of a particle.
Just as for the previously discussed case of electromagnetic fields, this contradiction
can be resolved by introducing the uncertainty principle. This principle has the form of
conditions restricting the range of coordinates and the range of momenta that can be
measurcd simultaneously for a particle. Uncertainties in the quantities l:!.p and l:!.r have
to satisfy the following inequalities:
l:!.P.T l:!.x ::=: ". (2.47)
Thus, if a particle is localized in a region of space of width 6x, the uncertainty in

the x-component of its momentum will be greater than or equal 10 hi l:!.x. The particle
described by the wavefunction of Eq. (2.43) has a certain momentum p = tlk, but is
completely delocalized in space.
Note that the phase of the wave <p in Eq. (2.43) depends on time. TIle angular frequency
ofthe oscillations of this phase is related to the energy oflhe particle E through Q = E lti.
The latter again coincides exactly with the energy of the \'lave quanta introduced by
Eq. (2.39). This kind of temporal phase dependence ex eiE / 11t and remains valid for
any complex system wldcr stationary conditions, including the condition of a constant
external field.
Another important fact is that the superposition principle discussed previously forclas-
sical waves is valid for particle waves. lllUs, typical wave phenomena such as interference
and diffraction should be observed for particles. One of the requirements necessary to
observe these effects is coherence ofthe waves participating in the superposition. In "par-
ticle language" this means, first of all, that particles should be monoenergetic. Indeed,
direct experiments with monoenergetic electrons have proved the occurrence of inter-
ference and diffraction of the electron waves. Sincc the famous 1927 Oavisson-Germcr
experiment on diffraction of electrons by metal crystals, numerous experiments confirm-
ing the wave nature of particles have been done. Recent (1989) experiments by Akira
Tonomura with diffraction of electrons involved repeating a double-slit experiment that
had been performed with light; see Fig. 2.7. In these experiments, direct confirmation of
wave-like properties of electrons was obtained. The experimental setup ofTonomura's
experiment is presented in Fig. 2.7(a). It consists of (i) an electron gun that emits, one
by one. electrons with high velocity; (ii) an electron biprism (electron splitter); (iii) a
detector of diffracted electrons; and (iv) a CCO (charge-coupled device) camera that
records and displays the positions of the registered electrons. Ten electrons per second
were emitted by the source. For the first several minutes the picture on the ceo screen
reflected a chaotic distribution of electrons. Gradually the build-up of registered electrons
Electron blprism
Dtltector ~;;;;;~l- •
__"':'--iH.ttM"H1
(0)
Agure 2.7 Tonomura's experiment. (a) A schematic diagrnm of the double-slit experiment.
(b) The build-up of interference fringes at various times: pictures on the monitor after (I)
10 electrons, (2) 200 electrons, (3) 6000 electrons, and (4) 140000 electrons had been detected.
Electrons were emitted at a rate of 10 per second. Used with pemlission, from A. Tonomura.
(2006). Double-slil Experimell1. (http://hqrd.hitachi.cojp/globaldoubleslil.cfm). figures 1 and 2.
~ Hitachi. Ltd. 1994,2006. All rights reserved.
produced an interference panern characteristic of experiments involving diffraction of

light in a two-slit experiment.
To estimate the particle wavelength and to understand the consequences of the uncer-
tainty principle, let us assume that a free electron moves with a velocity of about
t 01 cm s~ I. The mass of a free electron is 11/0 = 9.11 X 10- 28 g; thus, po = !/lOll =
9 X 10- 21 g cm S-I , ko = Po/ft = 8.7 x 106 CI11- 1, and the de Broglie wavelength of a
free electron is AO = 2Jr / ko = 7.2 x 10- 1 cm = 72 A. If we need to measure both the
position and the momentum of the electron, and we impose the limit of 10% accuracy
on the value of its momentum, i.e., I:1po = 9 X 10- 22 g cm S-l and I:1ko = 8.7 x 10 5
em-I, we cannot predict the position of this electron with an accuracy greater than
D.x = h/D.po = 2Jr/l:1ko = 7.2 x 10- 6 Clll = 720 A. This value is about ten times
greater than the wavelength of the electron!
According to our estimates, we see thai electron wavelengths have very small values.
For a material particle with a larger mass, the wavelength is even smaller. That is why in
most cases of ordinary life we do not observe wave-like behavior of particles. As we will
learn later, the wavelength of the elcctron in solids is tcn timcs larger than that in vacuum,
•
i.e., lip to 1000 A = 100 nm (or more). This is why, for an clectron in the situations
encountered in nanoelectronics, wave-like physical properties are its key attributes and
they determine the basic properties of this nanoworld.
2.5 Closing remarks
This chapter emphasizes a fundamental property of all existing marter, which may be
called wave-particle duality. Starting with an analysis of an arbitrary wave field (clastic
vibrations in solids, electromagnetic fields, etc.), we found particle-like behavior of this
2.6 Problems 29
wave field at small wave amplitudes and (or) spatial scales larger than the wavelength.
By analyzing particle motion, we have seen that at small spatial scales a particle can not
be characterized by exact coordinates and momentum (velocity) and behaves rather as
an extended wave-like object. Thus, wave properties and particle properties for the same
physical object arc compatible. This is the essence of wave-particle duality; indeed, this
is an underlying principle of central importance for nanophysics.
Contemporary microelectronics and optoelectronics exploit electric and light phenoll1~
ella, which are determined entirely by the properties of electrons in solids. Estimates of
the de Broglie wavelength ofelectrons in solids give us a value of about 10 nm and larger,
i.e., emerging nanoelectronics inevitably should be based on the wave mechanics.
For those who want to look deeper into wave-particle duality we recommend the
following textbooks:
R. P. Feynman, R. B. Leighton, and M. Sands, Lectures on Physics, vol. 3 (New York.

Addison-Wesley, 1964).
W. E. Lamb Jr. The Intelpretatioll a/Quantum Mechanics, edjted and annotated by J.
Mehra (Princeton, MA, Rinton Press, 200 I).
The problems presented below have two aims: to help the reader to attain a better
understanding of the definitions and principles stated in this chapter, and to illustrate
some unusual behavior of the electrons in solids.
In the following chapters, we will introduce basic definitions and equations of quan·
tum (wave) mechanics and analyze simple instructive examples that illustrate the main
quantitative and qual ilative features of nanophysics.
2.6 Problems
1. Formanysemiconductor materiaJs used in contemporary electronics, the relationship

between the momentum and energy of an electron is given by the implicit formula
L
2m 4
= £(1 + .E.).
EI,;
where m· is the so·called effective mass of the electron and I/EI,; is the so·caHed non·
parnbolieity parameter. The formula has two solutions for unknown E: for electrons
(E > 0) and another for the so·called holes (E < 0).
(a) Find both solutions for E.

(b) Derive the expression for \' for electrons only.
(c) Determine the electron velocity in free space and compare it with the expression
derived in (b).
As an example, consider GaAs, for which m· = 0.0671110 and E g = 1.42 eY.

Note: the expression for the kinetic energy of an electron in free space is
p'
£---
- 2mo'
2. Some metals and semiconductor materials instead of having an isotropic parabolic

energy dispersion, E = p2/(2",·), have anisotropic parabolic energy dispersion,
E = (_1)
m·..
PiP.
}
(2.48)
'J
i.e., for such cases "the electron mass" is no longer a scalar, but is instead a tensor. Lei
the reciprocal effective-mass tensor (Ij",t);j be
( 1) (
-I
/Ill
o-I ) . (2.49)
",t ij = 0 "'1
where i, j = x. y. Here, the parameters fill and /Ill are the so~cal1cd transverse and
longitudinal effective masses of the conduction electron. For this case, the dispersion
relation (2.6) is simplified to
, ,
E = Px + py . (2.50)
2m, 2ml
For the given momentum vector p = {ipisin (}, Iplcas (J)
(a) calculate the velocity" = dE jdp, and

(b) plot momentum veclors and velocity vectors corresponding to momentum vectors
with three values of B, (} = 30°. 45°, and 60°, and take Ipl = I. Consider n~Ge, for
which fill = 0.01911/0 and fill = 0.95"'0'
Notice that the directions of tile velocity vectors and momentum vectors do not coincidc
(i.e., they are not collinear).
3. Consider a particle of mass m, which moves along a single coordinate Assume =.

that the potential force is J = -fJz. Find the corresponding potential energy. Show that
the general solution to Eq. (2.8) is A sin(.Jf3/m 1 + </J), with A and ¢ being arbitrary
parameters. That is, the particle oscillates around the point z = 0 with the frequency
w = J f3/ 11/. Particular values of parameters A and ¢ determine the magnitude and the
phase of these oscillations.
4. To illustrate the method of generalization of the one-dimensional model, one can

consider a squarc (two~dimensional) lattice with identical atoms placed at the corners
of the squares. Each atom interacts with its nearest neighbors, i.e .• with four atoms, as
shown in Fig. 2.8. The equilibrium position of each atom is determined now by two
integer nlU11bers, say 11 and /II. The coordinates of the atoms can bc written as T II •1Il =
(XII' Ym) = {flO. ilia}. The atom with a given II and III has neighbors with coordinates
TII_I.m. TII+I. m • Tn.m-I, and Tn.m+l. If U II . m are displacements of the atoms from their
equilibrium positions, their new vector-coordinates are T'n.m = Tn.1Il + U n . m . Now the
displaccments U II .m and the forces are vectors. According to Hooke's law, the force
acting on the {II, m}th atom from its nearest neighbors is
2.6 Problems 31
II - I, III + II {II, m + I}
.-.
( u n.m ) . - ~:l(.' ..'. U n.n'
- ..' ..,, ..
{II. 111
{II-I,m} 11+ I,m}
11 - 1. 111 - I} II, m - I}
x•
Agure 2.8 A square two-dimensional lattice.
Calculate (/I.m, keeping only terms that arc linear with respect to displacements UIl.",_
Show that the force (/I,m has a form similar to Eq. (2. II). i.e ..
CIC. Write down the Newton equations for the displacements U",m_ Using the previously
discllssed replacement of the discrete system by the continuous medium, obtain a lWO-
dimensional wave equation. Find expressions for the density and the clastic modulus of
the two-dimensional clastic medium considered.
5. The simpiesl example of the interference effect is given by the superposition of two
sinusoidal waves (Ul = Bsinlf'lCr.r) and "2 = BsinfP2{r. f».
The resulting wave is a
sinusoidal wave of the form Umt = Bmt sinlP](r./).
(a) Find 'P,(r. I).

(b) Find 8 m•.
(c) Forthe spccificcaseof'P,(r. I) = =
Ii>,(r) + WI and'P2(r. I) liJ,(r) + wI, find 'PJ(r. I)
and Bmh and discuss the differences between your answers for this specific case and
answers (3) and (b) obtained before.
6. In addition to the solutions analyzed in Section 2.2. the wave equation (2.14)
has an infinite number of solutions of other forms. Let u(=) be an arbitrary func-
tion that satisfies Eq. (2.14), for which first and second derivatives du/d=, and
d 2 u/d: 2 can be calculated. Show that functions u(= ± Sf) are the solutions to the wave
equation.
7. Using the energy density equation
II = p(au)'
2 at
+ (au)'
ar
A
2
and the equation for a standing wave
u = Bsin(qnz)cos(wllt).
where IBI = I em,s = 340 m S-l, P = 1.29 kg m- J , V = L.rLyL;, and L.t = L.I' = L; =
2 m.
(a) Calculate the energy density of the wave for n = I, 2, and 3.

(b) Calculate the total energy of the wave for n = 1,2, and 3.
8. Consider a wave field inside ofa box: ofdimcnsions Lx. L y , and L;. The walls of
the box are supposed to reflect waves (a "mirror" box). Find three-dimensional solutions
of Eq. (2.18). Calculate wavcnumhers and frequencies of the standing waves. Calculate
the total energy of the standing waves.
9. The superposition principle of Eq. (2.31) allows one to present any wave field as
a combination of plane waves. The function u(t. z) = Be-H=-.Jr)/Il.=r is a wave pulse
propagating through the medium with the velocity s. By using the Fourier transform of
II, present this wave field as a sum of plane waves.
10. Consider visible (yellow) light with the wavelength A = 600 11m (600 x 10- 9 m).
Assume a light wave propagating in free space with the intensity density I = I mW cm- 2
(10- 3 J S-l cm- I ). Calculate the electric field amplitude Fo (in units ofVcm-'). Find
the wavevector, photon momentum, and energy. Estimate the density of quanta Nq.b.
3 Wave mechanics
3.1 Introduction
In this chapterwc discuss the basic physical concepts and equations related to the behavior
of particles in the nanoworld. We introduce the Schrodingcr wave equation for particles
and determine the ways to calculate observable physical quantities. We find that, in wave
mechanics, the motion of a particle confined to a finite volume is always character-
ized by discrete values of the energy and standing-wave-like wavefunctions. i,e., such a
motion is quantized. While motion in an infinite space (i.e., free motion) is nol quantized
and is described by propagating waves, the energy of the particle is characterized by a
continuous range of values.
Keeping in mind the diverse variants of nanostructures, by using wave mechanics we
analyze some particular examples. which highlight important quantum properties of par-
ticles. Many of the examples analyzed can serve as the simplest models ofnanostructures
and will be exploited in following chapters to understand the fundamentals of processes
in nanoelectronics.
3.2 The Schriidinger wave equation
From the previous chapter, we conclude that nanosize physical systems are qUillltum-
mechanical systems, inasmuch as their sizes are comparable to typical de Broglie wave~
lengths of the particles composing these systems. In dealing with quantum-mechanical
systems, one aims at determining the wavefunction of a single particle or of the whole
system. As we will demonstrate in the subsequent discussion, knowledge of the wave-
function in quantum mechanics is sufficient to describe completely a particle or even
a system of particles. This means that, if we know the wavefunction of a system, we
can, in principle, calculate all macroscopic parameters that define the properties of the
system.
The wavefullctioll of a particle \(I satisfies the principal equation of quantum mechan-
ics, the time-dependent Schrodinger wave equation,
a'll •
iI, - - 'IN = O. (3.1 )
at
34 Wave mechanics
where, ft, the reduced Planck constant, tl = II j(2rr), was introduced earlier, and the
-
operator 11. is the Hamiltonian of the system:
~ ft2 'il 2
H = - + V(r). (3.2)
2m
In quantum mechanics, the Hamiltonian is an operator, in contrast to the case in classical
mechanics, where it is a function. Let us consider the consequences of the basic equation
of quantum mechanics, Eq. (3.1). The Hamiltonian operator in quantum mechanics is
constnlcted by using the classical form of the Hamiltonian ofEq. (2.9), where the particle
momentum p is replaced by the momentum operator Ii = -i ft ajar. Thus, the first term
in Eq. (3.2) is the operator of the kinetic energy, with
, a2 a
'1-=-+-+-
2 a2 (3.3)
ax 2 8y 2 az 2
being the Laplacian operator. On comparing Eqs. (3.1) and (3.2) for particles with
Eq. (1.18) for wave fields, we may notice that both equations have second derivatives with
respect 10 the spatial variable, r, and, correspondingly, first and second derivatives with
respect to time,/. Despite the latter difference, the solutions ofEq. (3.1) are expected to
be in wave-like form.
If the potential V(r) is time-independent, one can separate the dependences on the
time and spatial coordinates:
(3.4)
where w(r) is a complex function of the spatial coordinates only. The time-dependent
wavefunction, \II(r. r), is often called the llon-stariollmy wallefimctioll, while I/f(r) is
referred to as a statio"my lI'avefilJlctioll. By substituting Eq. (3.4) into Eq. (3.1), one
gets the time-independent Schrodjnger equation:
- ~''1' + V(r) ) y,(r) = £y,(r). (3.5)

( 2m
In Eqs. (3.4) and (3.5), E is the total energy ofa particle. One of the important properties
of solutions to Eq. (3.5) is the orthonormality of solutions I/fi(r) and 1ft j(r) corresponding
to different values of the energy, E; and Ej :
f y,;'(r)y,j{r)dr ex o/j. (3.6)
where
Olj = II.O. if i = j.
if i '" j.
TIle major task of quantum mechanics is to solve the Schrodinger wave equation.
Eq. (3.1).
As we have mentioned already, the wavenmction of a particle in free space
(V(r) = 0) has a plane-wave form \.II(r. t) = Aei/J,;r-nn (Eq. (2.43». On substiniling this
3.2 The Schrodinger wave equation 35
£
'1',(:)
------_.
Figure 3.1 Three types of solutions of the SchrOdingcr equation for a one·dimcnsional weB of
arbitrary form.
w3vefUIlCl'ion into the Schrodinger equation (3.1), one gets the relationship between the
electron wavcvector and its energy,
111 pI
£ = tIn = 2m (t<; + ~ + J?,) = 2m' (3.7)
which coincides with the classical relationship between the part"ide's momentum. P. and
ils energy, £.
In general, the exact value of the energy E characterizes the system only for the so-
called statiol1wy-stare case. when the potential energy and, therefore, the Hamiltonian
do nol depend on time. In addition, for a stationary situation, a determination of energy
requires an infinite time of observation (measurement). If this time, t1f, is finite, the
inaccuracy of a mea~urement of the energy, t1£, should satisfy the following inequality:
(3.8)
This is the uncertainty relation between the energy and the time. Inequalities (2.47) and
(3.8) constitute fundamental uncertainty relations in quantum physics.
Equation (3.5) has the form known as an eigenvalue equation. The energy. £, is its
eigenvalue and the wavefunction, !/t(r), is its eigenfunction. The eigenvalue £ may run
over discrete or continuum values, depending on the shape of the potential nUlction,
VCr), and the boundary conditions.
In order to illustrate both possible types of solutions of the Schrodinger equation
and energy states, as well as to clarify the tasks which arise, let us consider the one-
dimensional problem for a system with potential energy VCr) = V(z) shown in Fig. 3.1.
Here the vertical axis depicts the energy £, and z represents only one space coordinate.
The potential has a negative minimum at z = 0, tends to zero as z ~ -00, and saturates
to a finite value V00 at z ~ 00. This potential is the most general form ofa potential well.
At this point. a short qualitative discussion serves to emphasize that the boundary con-
ditions define the type of solution. These solutions will be obtained and discussed in more
36 Wave mechanics
detail in due course. Among the possible solutions ofthe Schrodinger equation (3.5) with
the chosen potential, there can exist solutions with negative energy E = E i < O. The
wavefunction corresponding to energy E = E I < 0 is represented in Fig. 3.1 by curve 1.
One of the peculiarities of solutions with negative energy is that the spatial region with
classically allowed motion, where the kinetic energy p2/(2",) = E - V(z) > 0, is cer-
tainly restricted. In the classically forbidden regions where V(z) > E, the 1ft-function
decays as Izi ---+ 00. The states of particles, like t.hose just described, are the so-called
bOllnd states and they are characterized by a discrete energy spectrum. Consider next the
possible solutions for the energy range 0 < £ < V~ as shown by the line labeled £2 in
Fig. 3.1. These solutions exist for any values of £; they are finite as z ~ -00 and pene-
trate slightly into the barrier region V(z) > £ as shown by curve 2 in Fig. 3.1.ln the limit
z ~ -00, thesc solutions may bc represcnted as a sum of two wavcs traveling in oppo-
site directions: one wave is incident on, and the othcr is reflected from, the barrier. For
each energy in the rangc 0 .5 E .5 V00 there is only one solution satisfying the physical
requirements. For any energy £ > V00 there exist two independent solutions; see curve 3
corresponding to the energy £]. One solution may be chosen in the form of the wave
propagating from left to right. At z ~ 00 it has only one component. namely the wave
overcoming the barrier, and as z ~ -00 there is a superposition of incident and reflected
waves. It must be emphasized that the wave reflected from the barrier, when its energy
exceeds the height of the barrier, arises only in quantum physics. The other wavefunc-
tion can be choscn in the form of waves propagating from right 10 left. The latter is an
example of continuum energy spectra with energies £ > V00 when wave functions are
finite far away from the potential relief. These considerations dcmonstrate the impor-
tance of boundary conditions for Eq. (3.5): if the decay of wavefunctions (i.e., l/J -Jo 0 at
z -Jo ±oo) away from thc potential well is required, the discrete energies and the bound
states can be found and they are of principal interest; if the bOlUldary conditions cor-
respond to the incident wave, continuous energy spectra will be obtained. To conclude
the discussion of boundary conditions, let us mention that we often use potential ener-
gies with discontinuities. In this case, at the point of discontinuity of the potential, we
should require both continuity ofthe w{/I'efimction and cOl/til/l/ity of,he derivat;\'e ofthe
wavefimction w;th respect to coordinates.
Since Ihe Schrodinger equation is linear, it is clear that, if a function \J,I is a solution
of the equation, then any function of the form of constant x l.IJ is also a solution of the
same equation. To eliminate this ambiguity, we have to take into account the probabilistic
character of the wave function. Indeed, if a physical system is enclosed in a finite volume,
the actual probability of finding a particle in this volume must equal I, i.e.,
J 1'I«r. 1)1' dr = I. (3.9)
Equation (3.9) is called the normalization condition. II provides the condition needed to
determine the constant multiplicative factor of the wavefunction for the case ofa system
of finite size.
Ifan infinite volume is under consideration and the integ-ral of Eq. (3.9) does not exist.
there are other normalization conditions instead of Eq. (3.9). Consider, for example, the
3.2 The Schrodinger wave equation 37
so-called scattering problem, in which electrons come from infinity and are scattered
by a local potential. This case corresponds to energies £2 and E J for particle motion in
the potential sketched in Fig. 3.1. For this problem one can assrnne Ihe incident wave
to be a plane wave with a given amplitude A: \11(=.1) = Aeik=c- iE1 !", Then, the amount
of scattered waves will be proportional to the amplitude A, due to the linearity of the
Schrodinger equation. Very often the last condition is referred to as an initial condition
instead of as a boundary condition because we are dealing with the state before the
scattering and states after the scattering.
In an overview of quantum mechanics it is very useful to introduce the density of
particle flow. i. In classical physics, the density of flow is a vector, which specifies the
direction of particle flow and has a modulus equal to the number of particles crossing a
unit area perpendicular to the area per unit lime. In quantum mechanics. this quantity is
given by the following formula
i = -ih- ( ifI • VifI-ifiVifi .) . (310)

2m
For example, the density of flow of particles described by the plane wave of Eq. (2.43)
is straightforwardly found to be
i= hk lA1 ,= ~IAI' =vIAI'-

111 m
Average values of physical quantities

In light of tile probabilistic character of the description of quantum-mechanical systems,
we have to clarify how to determine the average values of quantities that characterize
such systems. The simplest case is the calculation of the average coordinate of a particle.
Indeed the absolute square of the normalized wavefunction gives the actual probabil-
ity per unit volume of finding a particle at a particular point in space, as studied in
Section 2.4. Hence, the average value of a particular coordinate, say I, is given by
(Z) = f ifI'ZifI dr = f zlifll' dr. (3.11 )
Thus, the integral over space gives the mean, or expectation value, of coordinate z. It
must be stressed again that the meaning of the expectation value is the average of a
number of measurements of the coordinate I carried out over an ensemble of identical
particles.
Equation (3.11) can be generalized to a more general form for the calculation of the
expectation value of any observable a:
(a) = (A) = f ifI'AifI dr. (3.12)
•
where the operator A is associated with the observable Q. From the definition of the
expectation value, Eq. (3.12), one can see that, if 1/1(1 is an eigenfunction of the operator
•
A and corresponds to a certain eigenvalue a,
(3.13)
38 Wave mechanics
then the expectation value of the physical observable a coincides with the eigenvalue:
(a) = a. For example. iflhe wavefullclion is the solution of the Schrodinger equation
(3.5), one may calculate the mean energy, (E):
(3.14)
However, if a particle were in a state with no well-defined energy, say a state that is
characterized by the superposition of the solutions 1/1 /:""
(3.15)
one would obtain the average energy in the form
(3.16)
Here, we take into account the orthogonalization and normalization conditions of

Eqs. (3.6) and (3.9).
It is necessary to explain the differences between the two cases given by Eqs. (3.14)
and (3.16). The first case is related 10 a system characterized by a wave function, which is
• • •
an eigenfunction of the operator A; in this particular case A = 1t. The second case cor-
responds to the situation described by a superposition of the eigenfunctions of the same
operator. Measurements of the value of the energy for the first case would reproducibly
give the same result, E. In the second casc, the measurements would give us differcnt
probabilistic results: cnergies Ej will be measured with their probabilitics lei 12 , and only
their average, (E), remains the same.
Importantly, by using the Schrodinger equation. straightforward calculations ofderiva-
tives of the average vector-coordinate (r) with respect to time show that
d'
m -,(r) = - (dV(r») .
dt- dr
i.e., the classical Newton equation (2.8) is recovered in terms of expcetation values
obtained using wave mechanics.
Thus, the Schrodinger equation describes the evolution of the wavefunction of the
quantum-mechanical system of particles. Its solution with proper boundary orland initial
conditions gives us all the information necessary to calculate macroscopic parameters
of the physical system and device operation being analyzed.
3.3 Wave mechanics of particles: selected examples
The main principles of quantum mechanics have been discussed in Section 3.2. To
understand new peculiarities of particles arising due to the wave-particle duality, these
basic principles of wave mechanics are applied here in several instmctive examples.
111C simplest of thesc arc related to the so-called one-dimensional case in which only
one-dimensional solutions are considered.

3.3 Wave mechanics of particles 39
• • ·•• •
•••
V=oo: : V=oo
11=4
tlJr:: - - - - •• -------.---- ~
~
• ~
.-" ~
f!' ~
•
""" - - - - .------------- n=3 .--"0

..::""
"0
.--"
"'" ---- n=2
-------------- ">
o" 3:'"
-- - -
n=1
--------------
~~~~
-L 0 .l.. -L 0 .l..
2 2 2 2
Allure 3.2 Solutions of the SchrOdinger equation for a quantum well with infinite barriers.
Eigenenergies. E•• are defined as £. = n 2£1. where £1 = TI Z]flj(2mL 2).
Ofcoursc, we live in a "three-dimensional world:' Thus, wavefunctionsofrcal particles

are dependent on the three-dimensional coordinate vector r. However, often, the potential
energy is dependent on a single coordinale, or two coordinates. For such cases, there
exists a significant simplification of the wave-mechanical description. Indeed, let V
be a function of the =-coordinate only. Then, we can introduce the two-dimensional
wavevector k ll = Ik,,,. k,,} and write the w3vcfunction as
\11(1. r) = e1{k,x+J:,..r- n f)1/t(z). (3.17)
i.e .• the wavefunction is a plane wave propagating in the {x. Yl·plane with the amplitude
modulated in the z direction. By substituting Eq. (3.17) into Eq. (3.1), we find the one-
dimensional equation
t,' d'
+ V(z) - ) (3.18)
( -2
111
<1:'
~-
E l{t(z) = O.
where
h'k'
£=E- II (3.19)
2m
Here E. h2kM(2m), and £ can be interpreted as the total energy, the kinetic energy of
"frcc motion" along the x- and y-coordinates, and the energy associated with motion
in the potential V(=). respectively. Equations (3.17) and (3.18) demonstrate fortile case
being considered that the motion along the =-coordinate is independent of motion in the
Iwo other directions and can be analyzed in terms of one-dimensional motion.
A particle between two impenetrable walls

Initially. let us consider the one-dimensional case. when a particle is placed between two
impenetrable (rigid) walls at == ±L/2, as depicted in Fig. 3.1. Thus. L is the width of
40 Wave mechanics
a one-dimensional well with infinitely high potential barriers, namely V(z) = 0 inside
of the well and V(z) = 00 outside of tile weB:
for Izl < L/2.
V(z) = 1°'+00. for Izl > L/2.
(3.20)
The spatially dependent part of the wavefunction 1/I(z) satisfies the time-independent
Schrodingcr equation (3.18):
~1 d' )
-- - - E "'(z) = 0. (3.21)
( 2m dz2
In the general case, the solution to Eq. (3.2 I) is
t(z) = Ae IK : + Be- IK =. (3.22)
where we have introduced
K =
f# me
~
2' (3.23)
Impenetrable walls are viewed as boundaries that exclude the particle. Specifically, for
the case of impenetrable walls, the probability of finding the particle at z ~ -L/2 or
z > L/2 is zero, i.e.,
"'(-L/2) = "'(L/2) = 0. (3.24)
These are, thus, the boundary conditions for lhe wavcfullction of a particle in a quantum
well with impenetrable walls. By applying these boundary conditions to the wavefunction
(3.22), we obtain
Ae- iKL / 2 + Be iK 1./2 = 0 for z = -L/2,

Ae 'KL / 2 + BC- iK 1./2 = 0 for z = L/2. (3.25)
A nontrivial solution of this algebraic system of equations exists if, and only if the
fOllowing determinant is equal to zero:
e-tK L}2 elK L/2
elK1.}2 c-iK1.}2 = 0.
which results in
sin(K L) = ° or KL = Iff/. (3.26)
The latter equation determines the "eigenvalues," Kn :

1f
Ko = L'" ,,= ±I, ±2. ±3, ...; (3.27)
11= 0 is excluded because for Ko = 0 Eqs. (3.25) and (3.22) give 1/J(z) ;;;;; O. From
Eq. (3.23) and (3.27), we find possible energies of this particle:
h2 K 2 h2rr 2
en = n,,2 (3.28)
2m 21/1L2
By substituting Eq. (3.27) into Eq. (3.25), it is easy to check that at a given 11 the
relationship between coefficients A and 8 is 8 = _e bJll A = (-It+ J A. Thus, if II is an
odd integer number, we obtain symmetric wavefunctions from Eq. (3.22):
t/t,,(z) = L cos (''''Z)

V(2 L ' 1I=±1.±3.... (3.29)
I f II is an even integer, we obtain anti~symmetric wavefunctions:
t/t,,(z) = HC sin
Z
). II = ±2. ±4.... (330)
In these functions, the factor J2/ L arises as a result of the normalization condition:
1-~
~ 1t/t"(z)I' dz = 1. (331)
The integer II is called the quanwm lIumber. Actually, we see that the energies E" do
not depend on the sign of the quantum number n. The same is valid for the physically
important quantity IVtIl(z)1 2. Thus, we can use only, say, positive quantum numbers,
/I > O. Let us write down the four lowest states explicitly:
" = I. 1/11=
(2 cos (1<Z)
'II. L ;
11 = 2. Vt2 = T. sin (21<-)
V(2 --;- ;
1/ = 3. Vt3 = V(2 (hZ) ;

I. cos L
11 = 4. Vt4 = V(2
L sin (41<'
T ). (3.32)
These solutions allow one to draw some important conclusions.

First, thc energy spcctrum of the particle confined by a potential well is discretc. That
is, instead of a continuous change of energy - as is characteristic in classical physics -
a quantum particle placed in a well may have only certain discrete energy values. In
other words, the energy spectrum becomes qu{mri::ed, as shown in Fig. 3.2. Often, one
refers to this type of the spectrum as a set ofdiscrete energy levels. Interestingly, for the
case bcing considercd, the interlevel distances, EIJ+J - Ell' incrcase with the number of
levels /I.
Second, the lowest energy level (usually called thegroll/ld swte) is not zero; it is finite.
That is, the particle can not have zero energy! Actually, this is a direct consequcnce of
thc uncertainty principlc. Indeed, a particle placed in a well is localized in a space rcgion
of size L, i.e., the uncertainty !J.z is less than or equal to L. According to Eq. (2.47),
such a localization leads to an uncertainty in the momentum !J.p ~ !J.p;: ....... Ii / L. Then,
thc momcntum of thc particle can be estimated by p ~ !J.p, i.c., thc momentum is not
zero and there is a nOll-zero total energy.
Third, the wavcfunctions arc, in fact, standillg waves, and strictly integer numbers of
half-waves may exist between these impenetrable walls. This result is mathematically
42 Wave mechanics
V=Vb V=Vb
OJ"
0
1'3
VI ~----
.-eo"...
"<::
"
-0 ~
~----
.-"
~
N
@ 1'," 0
::> ~----
0
..=l.. • ...l..... z
2 ° 2
Agure 3.3 Eigenencrgics of a particle in a quantum well with finite potential barriers.
similar to the case of the quantization of standing elastic waves between rigid walls
depicted in Fig. :!.4.
A particle in a quantum well with finite potential barriers

Now we consider the case of particle confinement in a potential well with potential
barriers of finite height. We accept the following idealized form of the potential as
illustrated in Fig. 3.3:
for Izi :5 L/2.

V(z) = 1°·Vb. for Izi ~ L/2.
(3.33)
where Vb and L arc the heighl and the width of the well, respectively. In this case, since
the energy spectrum is quantized such a potential well is frequently referred to as a
qua"tum well w;thfi"ite potential barriers.
Using our classical-physics intuition, we may expect that a particle will be confined
in the well if its energy s < Vb. Let us focus just on this case. First, we have to solve
Eq. (3.18). Inside the well, V(=) = 0, and the solution is a simple combination of sine
and cosine functions, as in the previous case ofEqs. (3.29) and (3.30). So, it is convenient
to rewrite it in a form that combines them,
t/!(=) = C cos(k.=l + D sio(kw=). for 1=1 :5 L/2. (3.34)
where
k", = J2ms/'ftl. (3.35)
and C and D are arbitrary constants. Outside the well the solution has the form
for = ~ L/2.
(3.36)
for = :5 -L/2.
J
where Kb = -2",(£ - Vb )/1l1 . As a result of the symmetry of the problem, we can
choose either even or odd combinations in Eqs. (3.36) and (3.34). COllsequently, conti·
nuily ofthe wave function implies that A = B for the even solutions and A = - B for the
odd solutions. ow we have two constants for odd solutions and two for even solutions.
In particular, for even solutions we find
for 1=1 :5 L/2.

(3.37)
for Izl ::: L/2.
where the signs "-" and .. +'. correspond to positive and negative values ofz, respectively.
The next step in finding the solution is to match the wavefunctions and their derivatives
with respect 10 z at the points z = ±LJ2. For example, for even solutions we obtain from
Eq. (3.37) the system of algebraic equations
C eos(kwL/2) - A = O.
(3.38)
Ck w sin(kwL/2) - AKb = O.
This systcm of algebraic equations has solutions if the corresponding determinant is

equal to zero. So, for even and odd solutions we obtain
tan ( kwL)
- - =Kb
- (3.39)
2 kw '
eof;L) = -Z: (3.40)
These trigonometric equations can be solved numerically, but it is more instructive to

analyze them graphically. Let us transform them into the following equations:
eos(kwL/2) = ±kw/ ko. for .an(kwL /2) > O. (3.41 )

sin(kw L/2) = ±kw/ ko. for eot(kwL/2) < O. (3.42)
where ko = j2mVb/t,2. The signs .'+" and "-" in Eq. (3.41) are to be chosen when
values ofcos(kwL/2) are positive and when they are negative, respectively. The same is
valid for "+" and "-" regarding sin(kw L/2) in Eq. (3.42) for the odd solutions.
The left-, C, and right-, 'R, hand sides of Eqs. (3.41) and (3.42) can be displayed on
the same plot as the functions of k ....: see Fig. 3.4. The right-hand sidcs ofEqs. (3.41) and
(3.42) are linear functions with slope equal to kr;l. The left-hand side is a cosine or sine
function. Intersections of these two curves give us values k..... 11 for which our problem has
solutions satisfying all necessary conditions.
To analyze the results. we note that the problem is characterized by two independent
parameters: the height ofthewell, Vb, and the width, L. We can fix one oflhese parameters
and vary the other. Let us vary the height of the well Vb, i.e .• the parameter ko. In this case
the left-hand side of Eq. (3.41) and the corresponding curves C. in Fig. 3.4 do not change,
but the slope of the line, k.... / k o, is controlled by k o. We can see that at small ko (small
Vb), when the slope is large, as shown by the dashed-dotted line in Fig. 3.4, there is only
one solution kw. 1 corresponding to small ko. This first solution exists at any ko and gives
the first energy lcvel El = tl2.k~.d(2m). As ko incrcases a new energy Icvel emerges at
44 Wave mechanics
c.n ,
.,' 11=1
~
I
,,
,,
----
slope kOI
o k",.1 JL
L
2ir
L
Agure 3.4 A graphical solution for Eqs. (3.41) and (3.42). Solutions comspond 10 Ihe
inlcrsections of the curves. The dash-doned line corresponds to small KO, resulting in only one
solution ofEq. (3 41). The doned line corresponds 10 a crilical value of KO OIl which a second
level appears in the wcll as a result oflhe solution of Eq. (.1 41). The solid line corresponds to an
inlennediale value of KO leading to four solutions.
--""' 15
~
(a)
'0
""0 10 --------------
.D
(b)
"""'0
~
"
.D
~
5 --- --
Z
""
0 100 200 300 L (A)
Figure 3.5 The number of bound slales of a square well planed as (J function oflhe well
thickness: (a) Vb = 224 I1lcV, m = 0.0671110; (b) Vb = 150 meV; m = 0.411I0'
kw = k o = Jr /L. as is shown by the dotted line in Fig. 3.4, with energy slightly below
Vb, (£2 ~ Vb)· Whcn ko increases further, the first and second levels become deeper and
a third level occurs in the well, and so on. Indeed, new levels occur when the parameter
J(2m VbL2)/(Jr2f12) becomes an integer.
Figure 3.5 depicts the number of bound states as a function of well thickness for two
specific values of Vb and for two particle masses. The parameters chosen are characteristic
for artificial quantum wells fabricated on the basis of AIGaA GaAs materials (see the
•
next chapter). One can see that for "thin" quantum wells (L < 100 A) only a few energy
levels can occur.
The first example we discussed is related to an infinitely deep well, when k o ~ 00. In
this case, the slope ofthc linear function in Fig. 3.4 tends to Zero, solutions correspond
to k", = 7f n/ L, and the energy levels are given by Eq. (3.28) as discussed earlier.
Thus. as for the first example. for a quantum well with walls of finite height, we have
also found discrete energy levels. The lowest energy level (the ground state) is also
non-zero. However, the number of discrete levels is finite. Another important phe-
nomenon is the penetration of a particle under a barrier. Equation (3.36) describes
the wave function under a barrier. To explain this effect, consider the energy (2.9) for a
classical particle: E = p2/(2111) + V(z). Since ,} is always positive, particle motion is
possible in the spatial region with s - V(z) ~ O. For E: < Vb, this "classically allowed"
region coincides with the region "within" the quantum well. The motion of such a clas-
sical particle is always restricted to a finite interval of the coordinates and we never find
the classical particle outside of the region where its total energy is less than the poten-
tial energy: i.e., a classical particle treats any barrier as an impenetrable barrier. The
quantum-mechanical analysis shows that the wave functions are finite at any coordinate
and we can find a particle even in classically forbidden regions, or, as one says, under
'he barrier.
The effect of pcnetration of a particle under a barrier is the so-callcd , w/lle/illg effect. It
is principally a quantum-mechanical phenomenon. The probability of finding the particle
rapidly decreases as we move away from the classically allowed interval.
A confining potential with quadratic coordinate dependence

Now we will briefly analyze the very important case of particle confinement in a potential
with quadmtic coordinate dependence:
I d2 V I
V(z) = 2 c1z2 z2 = 2mw2z2. (3.43)
where we introduce intentionally specific notation for the second derivative of the potcn·
tial V. Indeed, in classical physics, the force corresponding to this potential is a lincar
function: fez) = -mw'l= and the solutions to Newton's equation (2.8) are functions
oscillating with frequency w; see Problem 3 of Chapter 2. This type of particle motion is
that of a harmonic oscillator. Moreover, the first non-zero term in the Taylor expansion
of any potential ncar its minimum is quadratic with respect to displacement from its
minimum as in Eq. (3.43). This is why the example of the harmonic oscillator has wide
applicability.
For the harmonic oscillator potential, the time-indepcndent Schrodinger equation
(3.18) can be rewritten as
Z
2m (
-d'll/!
dz z
+,.
f,·
<-
mwZz )
2
1/1 = O.
It is convenient to introduce the characteristic length. Zo = jfl/(mw). In dimensionless

coordinates, ~ = z/zo, the wave equation takes the simple form
d2~ +
d~-
(2<t,W _~2)1/1 = o. (3.44)
Consider first JjJ at very large coordinates: ~2 » 2Ej(tiW), when the second tcrm in
brackets dominates. Then, Eq. (3.44) reduces to
d2 l/J' 1
d~2 - ~-1/1 = O. (3.45)
46 Wave mechanics
5hill
e,=
classical
turning point
11=1
hill
z
Agure 3.6 Eigenenergies and eigenfunctions ora panicle in a parabolic quantum well (harmonic
oscillator).
The asymmetric solution of the laner equation is proportional to e±l;~/2. Taking into
account that the solutions should be finite at I~ I -+ 00, we can look for a wave function
that will be a solution ofEq. (3.44) for all values of~ in the form
1/1(~) = .-"1' x p(n.
It can be shown that, to satisfy the normalization condilion (3.31) for wavefunctions,
the unknown function P(~) should be a polynomial ofa special form, which is called a
Hermite polynomial. This is possible only at certain values of the energy:
E, = tlW(1I + ~). II = O. I. 2. ... (3.46)
For each value of 11, the function P(~) can be calculated by using derivatives:
Returning to Ihe coordinate =, we presenl the thrcc lowesl states as

mw '/' -:lII2:~)
II = O. E. =
I
-flCu
2 . 1/1.(=) = (- ) • • :
,,~
'/' -
mw _ ..1,..1
n = I. 1/1,(=) =
( ) -
Jr 11
-.-. 1(_.,;
=0
(3.47)
II = 2.
5
- = -flW
E,
2 .
1/1,(=)=
( ) -
4Jlti
")
"'W '/' (-1+2, ._..1
.-./(-.,.
=0
,._
The quadratic polential, energy positions, and wave functions arc shown in Fig. 3.6.
Importantly, in contrast to the previous examples, the interJevcl distance is strictly con-
stant and equals fICu. This case can be classified as the case ofequidistalll energy /f!l'e!s.
As in the previous example. the particle pcnetrates, i.e., tunnels under, a barrier of finite
value. For comparison, we indicate that the classically allowed region of motion for the
particle is restricted by the two crossings of the curve V(z) and thc line E = constant,
as illustrated by Fig. 3.6.
Concluding these cxamplcs of particles placed into confining potentials, we cmphasize
the fundamental significancc of Planck's constant, 11: all quantized energies and interlevel
distances are proportional to ti. Ifwc formally suppose that tl -l> 0, the effects of energy
quantization, tunneling, etc., will disappear. That is, this limit corresponds to classical
mcchanics.
Quantized electromagnetic waves as harmonic oscillators

The harmonic oscillator model has a number of applications. The most important is
quantization of electromagnetic waves. This is one more illustration of the fundamental
principle of wave-particle duality.
In Section 2.3, we described briefly the quantization of electromagnetic waves. The
crossover from a classical wave characterized by the wavevector q and polarization ~ to
its quantized analogue is based on introducing the energy quantum given by Eq. (2.39)
and the photon number N q ,(. The energy of the wave must increase linearly with Nq,~. In
other words, energy states of the field are equidistant,just as for the harmonic oscillator
studied above, This link between the quantized field and the harmonic oscillators is
developed as follows. According to quantum physics, thc electromagnetic field consists
of an infinite number of modes (waves), each of which is characterized by a wavevector,
q, and a specific polarization, ~. Each mode {q. ~) may be described in terms of a
harmonic oscillator of frequency w = cq lj"i. Here, c and f:. are the speed of light and
the dielectric constant, respectively. Correspondingly, the energy separation between
levels of this quantum-mechanical oscillator is tiw = (tlcj../i)q (see Eq. (2.41)). This
oscillator may be in the non-excited state, which manifests the ground-state or zero-
point vibrations of the elcctromagnetic field. The existence of this zero-point energy is a
purely quantum-mechanical phenomenon, as manifested by the finite lowest cnergy for
particles in a confined potential. The oscillator may be excitcd to some higher cnergy
level. Let the integer N'I'~ be a quantum number of this level; then, the energy of the
electromagnetic field associated with the oscillator in mode {q.~} is
W V = (N
q., q ., + ~ )tJw. (3.48)
where Wq.~ is the energy density of the mode and V is the volumc of the system.
Equation (3.48) is the quantum analogue of the energy density given by thc classi-
cal equation (2.35). The excited-level number, Nq,~, is interpreted as the number of
quanta, or the number of photons in the mode under consideration. At large pho-
ton numbers, Nv . » ~, we immediately obtain the classical relationships given in
Table 2.1.
The lise of such an interpretation of wave-field quantization allows one to explain
all known effects of the interaction of electromagnetic fields with ma"er. For example,
48 Wave mechanics
kY
E/
"1>1 k.,
Flgure3.7 An example of two-dimensional subbands. The subbands are represented by the
paraboloids in 1£. kx • ky l·space; £1, £2, and E] indicate the botlolns of subbands.
the existence of zero-point vibrations completely explains spontaneous emission from

atoms, an effect that docs not exist in classical physics.
Low-dimensional subbands and "low-dimensional" systems

Now we return 10 the analysis of quantization effects related 10 particles. By generalizing
the examples of confined potentials previously considered, we can conclude that the
energy spectrum for a potential confining a particle in one dimension is
E=E k
".
=£ t,' ('
+-k-+k-.
" 2111 ' Y
') (3.49)
The energy spectrum depends on three parameters: two continuous parameters. k.{ and
k. and the discrete quantum number, 11. Such a spectrum is interpreted as a set 0/ellergy
j"
slIbbaflds: a set ofparaboloids with minima at quantized energies Ell, as shown in Fig. 3.7.
According to Eq. (3.49), a particle characterized by a given quantum number II can move
freely only in the {x. y}-plane. We can say that the particle possesses only two degrees of
freedom. In such a case, the particle may be viewed as a particle that is free to move in h.vo
dimensions. Thc two-dimcnsional character of particles becomes apparent when energy
intervals between subbands arc large. This, in turn, is possible when the potential well has
a width in the nanometer regime. Such a "quantum well" with low-dimensional particles
(electrons) can be fabricated by using thin (tens of nanometers) semiconductor layers.
Simple examples are just thin suspended films or free-standing structures (depending on
the fabrication method), as illustrated by Fig. 3.8. Another example is a thin layer placed
between two other materials. To explain this example, we will define in Section 4.5 the
energy required to remove an electron from a semiconductor to a vacuum outside of the
semiconductor - the so-called electron affinity. If a thin film possesses a large electron
affinity and surrounding materials are of smaller affinities, the difference between the
electron affinities explains the formation of a potential well for the electrons. Figure 3.9
3.3 Wave mechanics Df particles 49
(a) Suspended thin film (h) Free-slanding thin film

Rgure 3.8 Examples of suspended and free-standing thin films.
L B
Figure 3,9 A lhrce-Iaycr structure where Inyers A create potcnlial barriers for electrons from
layer B.
illustrates a heterostructure system fOTming quantum wells, which are important for
nanocleclronics.
Quite similarly. confinement ofa particle in two directions leads to additional energy
quantization and leaves only one degree of freedom for particle-wave propagation. The
energy for this case is
(3.50)
where the quantized energy £"1.11: depends on two integer quantum numbers. nl and "2.
and motion along the propagation direction (assumed to be the :c direction) is charac-
terized by the onc-dirnensional wavevcctor k¥. Thus, the spectrum consists of a set of
one-dimensional subbands as shown in Fig. 3.10 (compare the imaginary slices of the
parabaloids of Fig. 37 at fixed values ofk.,. with the parabolas of Fig. 3.10). Such a quan·
tized panicle is free in one dimension and the corresponding anificial structure is called
a quantum wire. Currently, there exist several technological methods for the fabrication
50 Wave mechanics
E"
Ell
Agu~ 3.10 An cxample of one-dimensional subbands. The subbands are presenled by the
parabolas in (E, k" I-space; Ell, Ell, and EIJ indicate the bottoms of subbands.
Flgure 3.11 A scanning electron microscope image of free-standing loP quantum wires. Reprinted
with permission from Thomas M<irtensson, Patrick Carlberg, ef al., ,; anowirc arrays defined by
nanoimprintlithograph," Nallo Lell., 4 (4). 699-702 (2004). American Chemical Society.
of quantum wires. For illustration, in Fig. 3.11, an array of frcc-standing quantum wires
made from InP is presented.
In the following sections we will show that transformation of spectra from three-
dimensional to low-dimensional energy subbands drastically changes the behavior of
the panicles.
Quantum-box, dot, and ·zero-dimensional" systems

So far, we have considered examples where a panicle is confined in one or two directions.
This leads to quantization of the electron spectrum. resulting in two--dimensional or one-
dimensional energy subbands. But there still is at least one direction for free propagation
of the panicle along the barriers of confining potentials. The advances in semiconductor
/17 L.
/ V/L, y
x
Figure 3.12 A sketch of a quantum box, L~ x Ll' X L:, embedded in a matrix.
technology allow one to go further and fabricate heterostructure-based structures, in

which all existing degrees of freedom of electron propagation arc quantized. These so-
called qua"tum dots, or quantum boxes, or zero-dimensional systems, arc much like
artificial atoms, and they exhibit extremely interesting behavior.
To consider the energy spectrum of a zero-dimensional system, we have to study
the SchrOdinger equation (3.5) with a confining potential that is a function of all three
coordinates and that confines the electron in all three directions. The simplest case is the
quantum box in the form ofa parallelepiped with impenetrable walls. The corresponding
potential, V(x. y. z), is
inside of the box,
V(x.y.z) = 1°·
1+00. outside of the box,
(3.51 )
where the box is restricted by the conditions 0 < x.:5 Lx, 0 <y ~ Ly,andO::: z ~ L:;
see Fig. 3.12. Using the result's of the one-dimensional analysis discussed previously,
one can write down the solutions of the Schrodinger equation for a box:
ftl7f2 (112I n1, ,,2)

J
= I. 2.3 .... ,
£11""2'''3 = 2
"'0
L'
.l
+ L'
Y
+ L' .
:
111.112.11]
8 Sill
L.,L)'L:
. ("XIII)
Lx
. ("YII')
L)'
.
Sin("ZII3)
L:
Sin . (3.52)
Offundamental importance is the fact that £"1. 112. 11 ) is the total electron energy, in con-
trast to the previous cases, where the solution for the bound states in a quantum well and
quantum wire gave us only the energy spectrum associated with thc transverse confine-
ment. Another uniquc feature is the presence of three discrete quantum numbers, "I, "2,
and,,], resulting straightforwardly from the existence of three directions of quant ization.
Thus, we obtain three-fold discrete energy levels and wave functions localized in all three
dimensions of the quantum box. It is important to stress here that the elimination of a
degree of freedom brings about a quantum number that substitutes for a continuous vari-
able, i.e., a component of momentum (or wavevector) corresponding to that particular
degree offreedom. In a quantum dot ofa parallelepipedal shape we have three quantum
52 Wave mechanics
numbers, Ill, "2, and "3, that substitute for the three components of the wavevector, k:
k." ky. and k;.
The discrete spectrum in a quantum box and the lack of free propagation of a particle
in any direction arc the main features distinguishing quanhlm boxes from quanhllll wells
and quantum wires. As is well known. these features are typical for atomic systems as
well.
The similarity with atoms is seen from another example of a potential- the spherical
dot. In this case, the potential can be modelled as
0. for l' < R,
V(r) = (3.53)
1+00, forI' ~ R.
It is instructive to compare this result with Eq. (3.20). We assume that the center of the
dot coincides with the center of coordinates; I' is the magnitude of the radius· vector, r.
and R is the radius of the spherical dot.
As we have found above, a quantum number (continuous or discrete) can be attributed
to each of the degrees offreedom of the particle. For a problem with spherical symmetry,
like that given by Eq. (3.53), it is convenient to introduce the following three degrees
of freedom: radial motion (along the vector coordinate) and two rotations, which can
be described by two polar angles, () and f/J. As in the case of a parallelepiped-shaped
quanhlm dot, the three degrees offreedom will bring abollt three quantum numbers. One,
associated with the motion along the radius, 1', substitutes for the wavevector component
kr , and the two other quantum numbers. associated with rotations. I and 111, substitute
for the wavevector components k(j and k4J.
The electron wavefunction in a spherical quantum dot can be found in the form
"',.I.m(r. e. <1» = "R,(r)Y,m(e. <1». (3.54)
where 'R1I(r) is the radial function and Ylm(O. ¢) are the s<rcalled spherical [1I11ctions,
which are angle·dependent. We will consider the simplest case of a wavefunction with
spherical symmetry, which corresponds to roo = I. For spherical symmetry both quan-
tum numbers associated with rotations are equal to zero: I = 111 = 0; this corresponds to
ko = kIP = 0, i.e., the wavevector, k, is parallel to the radius, r. On rewriting the radial
function as 'Rn(r) = XIl(r)/,. we find the Schrodinger equation for Xn(r) to be similar to
that in Cartesian coordinatcs:
h' a' ) x,(r) = 0.

( -2m
- - -2 - E
ar
Thus, as a result of the spherical symmetry the problem reduces formally to a one-
dimcnsional equal ion. For the case in which there is an infinitely high barrier surrounding
the quantum dot, the wavcfunction vanishes for I' > R and it follows that 'Rn(r) = 0 for
,. > R. One can obtain the solutions of this equation as
A sin(kwr)
'Rn (,. ) = . (3.55)
r
where k w = J2mo£ ItI at I' < R. 'R1I(r) = 0 at r > R, and A is an arbitrary constant. The
condition that 'Rn(r) = 0 for I' > R leads to the result that sin(kwR) = O. This equation
. .I--.". '~-F-(
•
-L 0 L
2. 2:
Rgure 3.13 Transmission of electrons through the barrier region.
is the same as Eq. (3.26) but with K.L replaced by kv. R. So, kv. is quantized: kv.' = if " ' R
(compare this wilh Eq. (3.27) and its analogy with Eq. (3.28» One can obtain
fi!;r"l1l2
£.00=
.. ,. ,. "=1.2 ..... (3.56)
un R-
where zero indexes show that the values of I and m arc equal to zero. This series oflevels
is the same as for the onc-dimensional quantum well with infinitely high barriers.
Quantum reflection, transmission and tunneling effects

Now we consider quantum effects arising for continuous energy spectra of particles.
Consider the simplest potential in the form of a rectangular barrier, as shown in
Fig. 3.13:
V(z) = I
Vb.
O.
fo< Izi :5 Lj2.
fo' Izl > Lj2.
(3.57)
where Vb. the barrier height, is greater than zero, and L is the barrier width. I.n such a
potential, the particle is not confined. The problem should be formulated as a scattering
=
problem: an incident particle moves, for example. from the left (from -+ -(0), then it
can be reflected back from the barrier or transmitted through the barrier into the region
= > L/2. The main goal of the analysis is to find the probabilities of reflection and
transmission. It is convenient to define these quantities as the ratios of reflected, i" and
transmiltcd, it. particle fluxes to the incident flux i m :
I, I,
R(e) = -. and T(e) = .,--. (3.58)
m I,.
'
The fluxes should be calculated according to the definition of Eq. (3.10).
There exist two distinct energy intervals, such that the behavior of the particle is
expected to exhibit different characteristics depending on which energy interval the
particle's energy is in: 0 < e < Vb and £ > Vb. In classical physics, for the first energy
interval, obviously the reflection coefficient is R = I, while the transmission is T = 0;
for the second energy interval R = 0 and T = I.
54 Wave mechanics
In wave mechanics, we have to solve Eq. (3.1 ) with the potential (3.57). The solutions
are
eik: + re-ik: . z :5 -LI2.
¢rez) = lie-it: + belt:. -L12 :5 z :5 L12. (3.59)
re ik : . z :: L12.
To the left of the barrier we introduce the incident \vave with a unit magnitude, eik: and
a reflected wave, re- Ik:. To the right of the barrierthere is only a transmitted wave, re ik:.
Then, the exponential factors are
k = ,j2m oE and
II
For E: < Vb, I( is a real number, whereas for E: > Vb, I( is an imaginary number. The
parameters r,ll. b, and r are still arbitrary functions of E, which we find by matching
these solutions and their derivatives at the walls of the barrier, z = ±L/2. The fluxes at
z -+ ±oo can be calculated in tcrms of,. and r: iin = I. i r = Ir1 2, and i l = It 12 . Omitting
the procedure for calculation of,. and 1, we write down the obvious relationship
R+T=1.
and the h'lO different results for the two cnergy intervals arc as follows:
1
T = ----:,-----:-;--- (3.60)
1+ (k' +K')' sinh'(2KL)
2kK
I
T = --.,.---,--,--- (3.61)
k' _K')' sin-(2KL)
,
1+
( 2kK
In the above formulas we used K2 = _1(2.
Let LIS consider first the case of an energy less than the barrier height, i.e.) Eq. (3.60).
Classical results can be obtained formally althe limit ofzero Planck's constant, f,. Indeed,
when ti -+ 0, we obtain I( -+ 00, sinh(21( L) -+ 00 and T -+ O. That is, no transmission
through the barrier occurs, as expected. In reality, this classical result is realized for
high and wide barriers (Vb. L -+ 00). However, at finite values of these parameters, we
always obtain a finitc probability of particle transmission through the barrier, which is,
as we discussed prcviously, the tunneling effect. For K L » I, sinh(2K L) = te21tl., the
second term in the denominator of Eq. (3.60) predominates and we can approximate the
formula as
16k 2 /(2
T :::::: e -41t I. (3.62)
(k' + K')'
Thus, the tunneling effect is determined primarily by the exponential factor. According
to Eq. (3.59) for the wavefunction, the probability of finding the particle under the barrier
is also exponentially small.
For the second case, corresponding to an energy greater than the barrier height, when
classical physics gives strictly T = L from the qua.nttlln·mechanical equation (3.61), we
3.4 Atoms and atomic orbitals 55
1.0 1.0
"7 (a) 0.• V~ (b)
0.• V
0.8
0.8
0.7
I,IOA·I 0.7
IIV 20A
0.6 0.6
I• ·1
>u
h O.5 ..,
0
h 0.5 >u
0.' 0.' ~
0
0.3 0.3
0.2 0.2
0.1
0.0
..... Barrier height 0.1
0.0
k:::::.. Barrier height
0.0 0.3 0.6 0.• 1.2 1.5 1.8 2.1 2.' 0.0 0.3 0.6 0.• 1.2 1.5 1.8 2.1 2.'
I: (eV) I: (eV)
Figure 3.14 Dependences of the transmission coefficient, T, on electron incident energy, €, for
• •
various thicknesses of the barrier, L: (a) L = 10 A and (b) L = 20 A. The height of the barrier
Vb = 0.3 eV.
find that Treaches I only at sin(2K L) = 0, i.e., at certain "resonance" energies, when
2K L = 1r n, with II beingan integer. Otherwise, T < I with minima at 2K L = (n + 1)1r .
All of this means that there exists a reflection of the particle even at large energy.
Figures 3.14(a) and 3.14(b) illustrate the transmission coefficient fort he barrier shown in
the inset. Both energy intervals, 0 < e < Vb and e > Vb, arc presented. The transmission
coefficient for £ < Vb is finjtc due to tunneling, but it is small; whereas for £ > Vb the
coefficient is close to I. Oscillations in T with energy arc explained by another quantum
effect - "overbarrier reflection" of the particles.
Concluding this analysis, \vc emphasize the great significance of quantum effects
in continuous energy spectra, and that these effects determine the basic electrical and
optical properties of nanostructures.
3.4 Atoms and atomic orbitals
It is instructive to compare the simple model of a spherically symmetric potential of

Eq. (3.53) and the simplest hydrogen atom. This atom consists of the positive proton
(nucleus) and the negatively charged electron, which interact according to the Coulomb
law:
I e2
V(r) = - - (3.63)
4Jffo r
where fO is the permittivity of free space, e is the elementary electrical charge, and r is
the distance between proton and electron. The negative sign in Eq. (3.63) indicates that
the electron and the proton are attracted to each other.
The potential of Eq. (3.63), shown in Fig. 3.15, has some similarities to that of a
quantum dot with impenetrable walls. This similarity allows us to suppose that the energy
of the electron in the hydrogen atom would be quantized. There are two major differences
56 Wave mechanics
r(A)
-20 -15 -10 -5 0 5 '0 '5 20
::ll ;:: 11-3
11=2
-5
;;-
~
"
'"
- 10
n=1
-'5
Agure 3.15 Eigenenergies ofan elef:tron in a hydrogcn·alom potential well.
between these potentials: (i) the profiles ofthe potentials (compare Eqs. (3.53)and (3.63»
and (ii) the reference points for zero energy. Namely, the zero energy was located in the
quantum dot and the electron should have infinite energy to escape into the potential
walls, whereas in the hydrogen atom the energy of a free electron is taken to be zero at
r = ±oo and the coer!,')' at r = 0 is equal to - . For further analysis of tile hydrogen-
atom model. the frcc-electron mass, "'0.
is used in this section because the motion of
nuclei can be neglected due to the fact that the mass of the nucleus is almost 2000 times
greater than the free-electron mass.
TIle ground state of the hydrogen atom may be described by the so-called radially
symmetric[lil/ction. 1/1("):
1{t(r) = (3.64)
where ro is the characteristic radius of the ground siale. It can be eSlimated as follows.
Let us evaluate the kinetic energy of an electron confined in the sphere of radius ro by
using thc unccrtaimy conditions of Eq. (2.0.17): p ~ til ro; then, the kinetic energy can be
evaluatcd as
(3.65)
Thus, the total electron energy, 'H, at r = ro is

p2 Ir 1)2
1i.(ro) = V(ro) + -2"10 = - 4JTEo -ro + 2moTo,. (3.66)
To find the ro that corresponds to the minimum oCthe total energy, 1f, we find the first
derivative of1f(ro) with respect to ro and equate the result to zero:
d1i.(ro) I t? 211 2
dro 4JTEO r02 - 2moT~ = O.
(3.67)
The solution of Eq. (3.67) gives the formula for 1'0:
(3.68)
which is called the Boh,. radil/s. Now we can find the energy of the ground state E I:
(3.69)
Numerical estimates give the following values for the energy and the radius of the
ground state of the hydrogen atom: £ I = -13.5 eV and ro = 0.529 A(I A = 10- 10 III =
0.1 mn).
An arbitrary atomic state is characterized by the following three quantum numbers.
n,/, and 11/:
principal number 11=1.2.3 ;
orbital number 1=0, I. 2 _, II - 1; or 1= s. p. d.... ;
magnetic number 11/ = O. ±l. ±2..... ±/.
The series of energy levels for the hydrogen atom is the following: Is, 2s, 2p, 35, 3p,
3d.... TIle ground-state w3vefunctiol1 of Eq. (3.64) corresponds to the case of n = I
and I = m = O. This wave function has the eigenenergy £ I of Eq. (3.69). The general
expression for the 11th energy level of the hydrogen atom is given by
(3.70)
Imponantly, the energy spectrum of the hydrogen atom depends 011 only one discrete
number II and docs not depend on the orbital and magnetic numbers I and Ill. This means
that some states of the atom can have the same energy. Tllis situation is called degeneracy
of the states.
Results from a number of detailed optical experiments have supponed this classifi-
cation of energy spectra of hydrogen atoms and confimled the energy values with high
accuracy. However, it turns out that nature is more complicated. Additional non-optical
experiments revealed a new property of quantum systems, which has no analogue in
classical physics.
Indeed, in classical physics a particle with angular momentum I can have a magnetic
dipole moment M = yl with the coefficient y called the gyromagnetic ratio. The latter
parameter is associated with the fundamental parameters of the electrons:
e
y=- . (3.71 )
2moc
where e, mo, and c arc the elementary charge, free-electron mass, and speed of light,
respectively. If an external magnetic field, B, is applied, an additional potential energy
58 Wave mechanics
of the particle in the magnetic field arises:
UM = -8M. (3.72)
Let the magnetic field be nonuniform, say it depends on the z-coordinatc. Then the force
acting on the particle will be
(3.73)
As a result, a beam of atoms with magnetic moments will split into components cor-
responding to different values of Af;. In special experiments, known as Stern-Gerlach
experiments. with hydrogen atoms in the ground s state, when f = m = 0 and thus
M= = 0, a splitting of the beam into lwO symmetrically deflected components has been
detected. It provided evidence of the existence of a dipole magnetic moment, which we
have not taken into account yet. It has been assumed that the electron has an intrinsic
angular momentum which is responsible for the results of the experiment. This intrin-
sic angular momenhlm S is called spin. The projection of the spin. S:, of an electron
in the magnetic field, B, takes on the half~integer values +! and -~. The spin effect
explains also the results of a number of other experiments with electrons. Moreover, it
has been found similarly that some spin characteristic can be attributed to any particle.
For example, photons - light quanta - have a spin equal to I.
Thus, the complete description ofan energy state of the hydrogen atom should include
additionally the spin number S:. Now the series of states including the spin degener-
acy can be written down as Is(2). 2s(2). 2p(6). 3s(2). 3p(6). 3d(IO).... The numbers in
brackets indicate thc degeneracies of the states. For example, the nomcnclature 2p(6)
describes six states of2p type (I = 1) with the same energy. but with different spins (±!)
and different magnetic numbers (m = -I. O. + 1).
Spin is a very important characteristic. It is responsible for a number of quantum
effects. Moreover, it determines the character of the population of different energy levels
by particles: in a system consisting of identical electrons only rwo (wirh differenr spins)
or/ewer elecrrons can beJound it! rhe same srare ar rite same rime. Since spin is included
among the quantum numbers, two electrons with opposite spins and all other quantum
numbers being the same should be considered as electrons in different quantum states.
TIle Pauli exclusion principle can be reformulated as follows: no two electrons can be
in the same state. This is the so-called Pal/Ii exclusion principle. This is one more quan-
tum principle. The Pauli exclusion principle is especially important for many-elecrrrH/
systems.
The classification of energy levels llsed for the hydrogen atoms may be applied to
many-electron atoms. In the case of many-electron atoms, we should take the nuclear
charge as Ze. where Z is the atomic number in the Periodic Table ofelements (Fig. 3.16),
and take into accoulltthe Pauli exclusion principle. The electronic structure of complex
atoms can be understood in terms of the filling up of energy levels. One can introduce
internal energy shells and external energy shells (or valence energy states). The latter
are filled up by the outermost valence electrons and play the major role in determining
the chemical behavior of complex atoms.
1 17
,
11
,-,
L1 .•
2
,
13
,
B
14
•
C
15
,
N
"
•
0
..
f
,
,11,
,
"
II,
'00'
"
Nt
.~, 'J1l12 Trnnsilion Metals • 10 " 12.01 1401 "00 '"00 20 I'
"
N.
l:!.W
Mg"
~4JO J • 5 , 7 , 9 10 11 12
"
,\1
~6 'Ill
"
S1
noo
"
P
)091
"S C1"
1206 IHS
M"
,,~
" " "V " Mn

"
, " • " " " " •
K
N
C. S, "
1'1 C, ", "
C. "
Nl C. Z. G. "
G. ," S, B, K.
l'Jlll ~oo ~~ 4Hl ~~ .,00 ~~ 5S .5 n9J ~~ 6B5 65-39
~" lHoI 14~2 n. ~oo
"'.
"
Rb "
S, " l\Ill
"V Z, Nb " 1', "
00 •
R, "
Rb
•
Pd "
A, "
Cd
•
I.
•
S. "
Sb "
1', "1 "
XI:'
,,,, ,., ,,,,
1147
"
1161
"" "" •• Iq'l
• " \If T." W" R,"
r.
QUI ,,~ 101 1
,
" " "
lIH
•
II••
"
HIT
"
Ill.
"
127&
" ..,"
Wi9 IlU
"
Co
112 q
"
f,
B. L,
lJ1 J
"
R.
IlSll
'"
L,
,. ,.,
ms
RI
,,,
Db
,.
III •
S,
I.td
10'
Bb
'",. IU2
'00' ,."
1\11
Po
19S I
". '"
A'
197.0
IIg
~.
III
1'1
~,
Pb
1072
PI
~,
P,
120901 ,~IO 01
R.
,~llO.
;:12JO) ,n60) 1~6~ II 1261 II ,~~2.1t '"

.2tolll 126-1lj 126S II [2M> II 12r.9 II 1272.1' 12nll
" • " 0,'

...." P," Nd" Pm" SmOO," E," Gd • " E, • Tm
• YO"
L.
llU
'" , ,.'
.
• n" P. "U Np
,.'
" P,• Am
'fU
• Cm
T' , , II•
, ,.'
• " CI• E." I'm Md
IJ10
'" No'"
"72
.,
I'n
.' . 1671 17)1
," .. , , , " , '" ' , " . ,,, , ... ,,,

A,
Figure 3.16 The Periodic Table or elements.
While the energy levels of the hydrogen atom exhibit some degree of degeneracy.
i.e., they arc independent of the angular momentum described by the orbital number I
and its projection described by the magnetic number /1/, in many-electron atoms, the
energies depend not only on the principal quantum number 11, but also on the orbital
number /. Therefore, the electrons can be subdivided into "subshells" corresponding to
different orbital numbers I. The spectroscopic notations for these subshells are the same
as above: s. p. d. f. ... For a givcn I there are 2/ + I values of m and two values of
the spin S: = ±~. Thus, the total number of electrons that can bc placed in subshelll is
2(21 + I). Periods ohhe Periodic Table of clemenIs are constructed according to filling
of the shells. Hydrogen, H, is the first element of the table. It has one electron occupying
the Is state, i.e., ( = O. The next atom is helium, He. It has two electrons, which occupy
the lowest s states in the same shell, but have opposite spins; this is referred to as the Is2
electron configuration. So, the Is shell is completely filled by two electrons.
The next period or the table begins with lithium, Li. It has three electrons, with the
third electron sitting on the 2s-type level. Thus, we may immediately represent the
electron configuration of Li as Is 2 2s l • After lithium follows beryllium, Be, with the
configuration Is2 2s 2 • In the next elcment boron, B, the fifth elcctron begins to fiJI up
the next subshell: Is 2 2s2 2 p l. For the p shell, wc have I = I and m = -1. 0, 1. Thus,
the valence electron in Be can take three valucs of the magnetic quantum number, m,
and two values of the spin. S:. In general, we may put six electrons in the p shell, and
so on. For nanoelcctronics the elements of group IV and groups III and V that con-
stitute semiconductor crystals arc important: silicon (Si, whose atomic number. and
number of electrons, is 14), germanium (Ge, atomic number 32), gallium (Ga, atomic
60 Wave mechanics
Antibonding orbital lI'AB
(a) (b)
(c)
A B
Atomic orbitals lJIA and lI'B
with equal energies of atoms
A and B Lower level "'---
nuclcus nuclcus
Bonding orbital V'AB
FIgure 3.17 Formation or bonding and antibonding atomic orbitals (IWO atoms).
number 31), arsenic (As, atomic number 33), etc. Their electron configurations are as
follows:
group IV: Si: Is22s22p63s23p2. Ge: Is22s22p63s23p63d104s24p2;

group III: Ga: Is22s22p63s23p63d104s24pl; (3.74)
group V: As: 1s22s22p63s2 3p63dlo4s24p3.
One can see that, both for Si and for Ge. the outcr valcnce clcctron configurations arc
very similar (the same number of electrons in the same p state). Thus, we can expect that
their chemical and physical properties are also similar.
Atoms with completed subshells are stable and chemically inert. Examples are helium,
He, with the configuration 1s22s2; neon, Ne, with the configuration Is2 2s22 p6; and argon,
Ar, with the configuration Is22s22p63dlo4s24p6.
In general, atoms with not completely filled-up subs hells, that is. those which have
valence electrons, are expected to be chemically active. i.e., they can form chemical
bonds. Coupling beh....een atoms in molecules and solids is determined by the type of
thc electron wavefunctions. Often, these wavefunctions are called atomic orbitals. For
example, the wave function of a two-atom molecule should be a combination (a hybrid)
of two orbitals. Such a hybrid can be either a symmetric function, or an anti symmetric
one. Let this hybrid be a symmetric function (constructive interference oftile orbitals), as
shown in Fig. 3. I 7(c). Then, there is a finite probability of finding the electron between
the nuclei. The presence ofelectrons between two nuclei, A and B, attracts both nuclei and
keeps them together. The corresponding state is called a bonding state of two atoms and
the corresponding wavefunction, 1/JAB, is called the bo"ding orbillll. The antisynulletric
hybrid (destructive interference of the orbitals) does not bond atoms and is called an
a"tibollding state, see Fig. 3.17(b). The simplest example of the bonding effeet is given
by the two-atom hydrogen molecule: the bonding state constructed from s states of
the hydrogen atoms corresponds to an energy lower than the energy of hvo uncoupled
z z
(e)
y y
x
s orbital p)' orbital p,oroital
z
(a) (b) (d)
-- -
,, --'y"
,
y y ,, - y
,).,j-./
X
Pol" orbital
Figure 3.18 Atomic orbitals: (a) s, (b) P'" (c) p=, and (d) p.r.
hydrogen atoms. This results in a stable H2 molecule. The anlibonding state corresponds
to an energy greater than the energy of two free hydrogen atoms. In the anlibonding state
the atoms repel each other.
Actually, the formation of bonding and anlibonding slates of two hydrogen aloms is
a particular example of a morc general and very important result: when two identical
atoms approach each other closely, their energies change and, instead of a single atomic
(degenerate) energy level, two levels arise, as illustrated in Figs. 3. I 7(b) and 3.17(c). This
is the so-called energy spliflillg. Thus, bonding and antibonding orbitals correspond to
these two levels. Three atoms drawn together are characterized by three close levels, etc.
Now, consider in detail atoms of group IV. As presented above, for these atoms the
outer shell has four electrons: two in s states and two in p states, i.e., their configuration
can be thought as a core plus the S2 p2 shell. Thus, atoms of group IV can form four bonds
with other atoms in molecules or solids (onc bond for one clectron). The wavcfunctions
of these outer-shell statcs have very different spatial configu.rations. Figure 3.18 illus-
tratcs the wavcfullctions (orbitals) for an s state and for p states with different angular
momentum projections m. Thus, the s orbital is spherically symmetric without any angu·
lar dependence. The 3p orbitals are anisotropic and can be considered as "perpendicu.lar"
to each other. The hybrid orbitals forming bonds with other atoms have two electrons
with opposite spins. In fact, because of interaction between electrons the sand p orbitals
overlap significantly and realistic bonds are formed as linear combinations of sand p
orbitals, which are called Sp3 hybridized orbitals. Importantly. when the three directed p
orbitals are hybridized with the s orbital, the negative parts of the p orbitals are almost
cancelled out, so that mainly positive parts rcmain, as showll in Fig. 3.19(a). Thus. the
hybridized orbitals prescnt "directed" bonds in space, as shown in Fig. 3.19(b). These
directed bonds of atoms are responsible for the particular crystal slruchlres of elements
62 Wave mechanics
-
,'+--
-
-,r--
,, y
x
s orbital
--+
II
sp' orbital
x (a) (b)
PI orbital
Rgure 3.19 Formation of hybridized Sp3 orbitals (a) and four spJ hybridized orbitals in Si (b).
in group IV. In the next chapler, we will analyze these crystals using the definitions
discussed above.
3.5 Closing remarks
In this chapter we have presented basic definitions and equations of quantum (wave)
mechanics. We found that two cases of particle mol ion should be distinguished: (i)
quantized motion always occurs with in a finite volume (for example, in a potential
well); (ii) free motion occurs in 3n infinite space (practically, in a large volume). We
discllssed the difTerences in particle behavior between these IWO cases.
We analyzed simple instructive examples thai illustrate the main qualitative fea-
tures of quantum physics. Then. we briefly discussed the electronic properties of
atoms, which we should use for the analysis of nanoclcctronic systems. Many of the
basic quantum-mechanical concepts introduced above will be applied in the follow-
ing chaplers to snldy semiconductor nanostructures and the principles of operation of
nanodevices.
For those who want to look deeper into the basic principles of quantum physics we
recommend the following textbooks:
L. Schiff. Qllolltllm Mechanics ( ew York, McGraw-Hili, 1968).
D. Saxon, Elementary Qllollfllm Mechanics (San Francisco. CA, Holden-Day, Inc.,
1968).
R. P. Fcynman. R. B. Leighton, and M. Sands, Lectllres 011 PhysicS, vol. 3 (New York,
Addison-Wesley. 1964).
The problems presented below illustrate the main definitions given in this chapter and
allow one to practice with the simplest quantum-mechanical examples.
3.6 Problems 63
3.6 Problems
1. Suppose that a particle is placed into a region of space of volume V. Using the
normalization condition of Eq. (3.9), find the amplitude A of the plane wave (2.43).
2. Assume that the wave function is a superposition of two plane \Wlves of the same
energy: I.IJ = 1/11 + 1/12 with 1/11.2 = A1.2C l(k l ..lr-nt J• Calculate the probability density of
finding a particle at the point r, i.e., calculate )\IJ(r)1 2 • Discuss the interference effect and
its dependence on the wavevectors k 1 and k 2 •
3. Assume that an electron with mass 11I0 = 9.8 x 10- 31 kg is placed in a quantum well
with two impenetrable walls and that the djstance between the walls is L = 10- 6 CIll.
Calculate the three lowest subband bottom energies: E\. £2, and EJ. For these stationary
slates, find the probability density of finding the electron at the middle of the well at
z =0.
4. Consider an electron, which is placed into a quantum well with width L = 10- 6 cm
and barrier height Vb = 300 meV Calculate the lowest energy level and compare your
result with results for the previous problem. Since the probability density for finding the
particle at a point z is 1l{t(z)1 2 , the probability of finding the particle in the classically
forbidden regions is
P= 2x 1 00
L{'
1\'I(z)I' <!z.
Use the solution for the lowest energy level to calculate P. Find the number of discrete
energy levels in such a quantum well.
5. The hydrogen molecule, H2, has the frequency of oscillations w = 8.2 x 10 14 S-I .
The reduced mass, which determines the relative displacement of hydrogen atoms, is
m = 0.84 x 10- 27 kg, which is half of the mass ofa hydrogen atom. Using the model
of a harmonic oscillator, estimate the characteristic displacement, zo, during vibrations
of this molecule.
6. In classical physics, for a system under equilibrium conditions, the equipartition

principle is valid. According to this principle the average energy of any oscillator is
keT, where k a = 1.38 x 10- 23 J K- 1 is Boltzmann's constant and T is the ambient
temperature. Calculate the energy of zero-point vibrations of an electromagnetic wave
of wavelength I0 ~m. Calculate the average number of photons, Nq.~, that will be excited
with this wavelength if the ambient temperature is equal to 500 K.
7. Consider two quantum objects, a cubic box and a spherical dot surrounded by impen-
etrable walls. Suppose that the volumes of the classically allowed regions arc equal.
Compare the ground-state levels for the box and the dot.
64 Wave mechanics
8. By using approximate formulae for the transmission coefficient, calculate the prob-
ability of tunneling of the conduction electron through an AlGaAs layer of thickness
•
20 A embedded within a GaAs matrix, for a barrier height equal to 0.3 e'J, with the
electron mass in GaAs III = 0.067"'0. where "'0 = 9.8 X 10- 31 kg is the mass of a free
electron.
9. One of the experimental facts which constitutes one of the fundamentals of wave
mechanics is the observation of discrete optical spectra (spectral lines) of cmission of
hydrogen atoms. Using the formulae for the encrgy levels of a hydrogen atom, En.
calculate a general expression for the emission frequencies. Find the wavelengths for
several lines of the so-called Lyman series, for which one of the states participating in
emission is the ground state 1: I ++ 2. I ++ 3. 1 ++ 4....
4 Materials for nanoelectronics
4.1 Introduction
After the previous introduction to the general properties of particles and waves on the
nanoscale, we shall now overview the basic materials which arc exploited in nanoelec·
tronics. As discussed in Chapter I, electronics and optoelectronics primarily exploit the
electrical and optical properties of solid-state materials. The simplest and most intuitive
classification of sol ids distinguishes bet\\leen dielectrics, i.e., non-conducting materials,
and metals, i.e., good conducting materials. Semiconductors occupy a place in between
these two classes: semiconductor materials are conducting and optically active materials
with electrical and optical properties varying over a wide range. Semiconductors are the
principal candidates for use in nanoelectronic structures because they allow great f1exi·
bility in the control ofthe electrical and optical properties and functions of nanoelectronic
devices.
The semiconductors exploited in microelectronics are, in general, crystalline materials.
Through proper regimes of growth, subsequent modifications and processing, doping by
impurities, etc., one can fabricate nanostructures and nanodevices starting from these
"bulk-like" materials.
Other physical objects that demonstrate promising properties for nanoelectronics were
discovered recently, for example carbon nanotubes. These wire-like and extended objects
are of a few nanometers in cross-section. They can be produced with good control of
their basic properties; in particular, they can be fabricated as either semiconductors or
metals. Various types of processing techniques have been shown to be viable for the
fabrication of electronic nanodevices from carbon nanotubes.
In this chapter, we consider various materials that have applications for nanoelectron-
ics. We start with the classification of dielectrics, semiconductors, and metals. Then.
we define electron energy spectra, which determine the basic properties of the electrons
in crystals. For nanoelectronics, a critical issue is the engineering of electron spectra,
which can be realized in heterostructures. Thus, we analyze basic types of semiconductor
heterostructures. Finally, we briefly describe organic semiconductors and carbon-based
nanomaterials, among them carbon nanolubes and such nano-objects as fullerenes.
-. -- - - ---- E,
-
-.--------
---------
E,
E,
-
<::
Energy level E,
--------- E,
<::
Energy le"el E~
Overlap
,~"e;
Large number of liloms
E,
< Energy level EI
(
(a) (b)
Figure 4.1 (a) Fonnation of all energy band. The ground energy level E J of a single atom evolves
into an energy band when many identical aloms arc interacting with each olher. (b) Energy
bands 1,2, and 3 that correspond 10 single-atom energy levels £., £2, and E l , respectively.
4.2 Semiconductors
In every solid, electrons can be chamclerized in terms of their energy levels. In crystals,
allowed electron energies typically have an energy band structure that may be understood
as follows. In Section 3.4, we pointed oul that. if two atoms, each with the same energy,
come close to each other, the composite two-atom systcm is characterized by two close
energy levels. Similarly, for a system of N atoms, every energy level of the isolated atom
is split into N closely spaced levels. This assembly of close levcls may be considered as
an energy band. Figure 4.1 (a) illustrates schcmatically the formation of such an energy
band from a single atomic Icvel. Since a single atom has a series of energy levels, the
electron energies in a crystal constitute a series of energy bands that may be separated
by energy gaps, or may overlap as illustrated by Fig. 4.1 (b). As soon as the energy bands
are formed, the electrons should be thought of as collectivized: they can no longer be
attributed to specific aloms, since the energy bands characterize the whole system of N
atoms.
A crucial point is the filling of the bands by electrons. We shall use the Pauli exclusion
principle, which we considered in the previous chaptcr. That is, no 1\,",0 elcctrons can be
in the same state. It is possible, as an example, for two electrons to be in the same energy
state, but these electrons must be in di.fferent spin states; thus the electrons are in facl in
different overall states. Under equilibrium conditions and at low ambient temperature,
the lowest energy levels should be populated. As we will see later, the most important
electrons are those in the upper populated bands. Then, principally, we obtain two possible
cases.
First, all bands are completely filled and the filled bands are separated from the upper
(empty) bands by an energy gap. This is the case illustrated by Figs. 4.2(a) and 4.2(b) for
dielectrics (insulators) with bandgaps £8 > 5 eVand for semiconductors with bandgaps
£8 < 5 eV, respectively. Actually, there is no difference bel\veen filling up energy baJlds
for a dielectric (insulator) and for a semiconductor. The difference is in the energy gap
4.2 semiconductors 67
(a)
Empty (b)
t Empty
•••••••••••••••••
('
•• ••• • • • •• •
• • • • • • • • • • •
• • • • • • • • • • • • • • • • •
• • • • • • • • • • • • • • • • •
(e)
(d)
Partially Filled
• 0000000000000000 • Empty
• 0000000000000000 •
• • ••••••••••••••••••
••••• • • ••• •••••••• Overlap
•••• • •••• •••• •••••
•• •• •• •• •• •• •• •• •• •• •• •• •• •• •• •• •• ••
•••• ••• •• ••••• • • • ••• • •
• • • •••
•• • •• •• •• •• •• • • • • •
• •• •• •• •• • ••••••• ••• • •••
• •• • ••• •••••••••••••••
•••• ••••••••• ••••• • • • • • • • • • • • • • • • • • •
Figure 4.2 Enclb'Y bands in the cases of a dielectric. a semiconductor. and a metal. (a) The case
of a dielectric: filled valence band and empty conduction band; E g > 5 e V. (b) The case of a
semiconductor: filled valence and empty conduction band (at low temperature); E g < 5 eY.
ee) Electrons in the partially filled band may gain energy from an electric field and as a result.
trdnsfcr to empty ("free") levels of higher cnerb')'. thus exhibiting electric conductivity.
Accordingly. a structure with a parlially filled band corresponds to a metal. (d) For oyerlapping
bands the available electrons fill slalcs in both bands.
between the upper filled band and the next empty band: for semiconductors this energy
gap is much smaller than it is for dielectrics, as is illustrated in Figs. 4.2(a) and 4.2(b).
This upper band is empty, at least allow temperature. Later, we will study the electronic
band structure of semiconductor materials in more detail.
Second, the upper bands that contain electrons are not completely filled as in
Figs. 4.2(c) and 4.2(d); this case corresponds 10 a metal. Indeed, to exhibit conduc·
tivity under an applied electric field, an electron should experience acceleration and a
gain in energy. That is, an electron should be able to gain a small amount of energy
and be transferred to close, but higher, energy levels. Ifall energy levels (a whole band)
arc already filled up, the electron can not participate in conduction processes. This is
the case of a dielectric (an insulator). In contrast, if the energy band is not completely
filled and there are empty energy levels available for the electrons, in an electric field
the electron can move to upper levels and gain energy in a process that corresponds to
electrical conductivity. TIlis is the case of a metal.
Semiconductors include clements from the central portion of the Periodic Table ofcle·
ments - columns II to VI, as shown in Table 4.1. At the center of Table 4.1 is silicon, Si,
the backbone material of modern electronics. Silicon plays the central role in electronics
Table 4.1 The central portion of the Periodic Table of elements
Group II Group III Group IV Group V Group VI
Be B C N 0
Mg Al Si P S
Zn Ga Ge As Se
Cd In Sn Sb Te
Hg TI Pb Bi Po
just as steel plays a dominant role in metallurgy. Below Si is germanium, Ge. Nowadays,
Ge is rarely used by itself; howevcr, Ge-Si alloys play an increasingly important role in
today's electronics tcchnology. Besides the elemcntal matcrials, contemporary elcctron-
ics also uses combinations of c1emcnts from group UI and group V, and combinations
of elements of group II and group VI, as well as some more complicated combinations.
These combinations are called compound semiconductors. By combining each element
from group III with N, P, As, Sb, and Bi from group V, 25 different III-V compounds can
be formed. The most widely used compound semiconductor is GaAs (gallium arsenide)
and all III-V semiconductors are used to fabricate so-called hererostrllctllres. A het-
erostnlcture is made of two different materials with a heterojunction boundary between
thelll. The specific choice of heterostructure depends on the application.
Two or more compounds may be used to form a/loys. A common cxample is aluminum
gallium arsenidc, AlxGal_xAs, whcre x is the fraction of group 1II sites in the crystal
occupied by AI atoms, and 1 - x is the fraction of group III sites occupied by Ga atoms.
I'fence, now we have, not just 2S discrete compounds, but a continuous range ofmaterials.
As with the III-V compounds, every element shown in column II may be used together
with every element in column VI to create II-VI compounds, and again, by combining
more than two of these elements, it is possible to create a continuous range of materials.
As a result. it is possible to make compositionally different IV-IV, III-V, and II-VI
compounds.
4.3 Crystallallices: bonding in crystals
We start with the definition of crystals. A cI)lstal is a solid in which the constituent
atoms are arranged in a certain periodic fashion. That is, one can introduce a basic
arrangement of atoms that is repeated throughout the entire solid. In other words, a crystal
is chamcterized by a strictly periodic internal structure. Not all solids arc crystals. In
Fig. 4.3, for comparison, we present a crystalline solid (a), a solid without any periodicity
(a so-called amorphous solid) (b), and a solid in which only small regions are of a
single-crystal material (a so-called polyclystalline solid) (c). As might be expected
crystalline materials can be the most perfect and controllable materials. Before studying
periodic arrangements of atoms in crystals, we shall discuss different types of bonding in
crystals.
4.3 Crystallalliees 69
••••
••••••• • ••
••••• •
••••••••
•••••••• •• •
•••• •••
• •
•••••• •• • • •
••••••
••••••• •• • ••
• ••
••••••• ••• ••••
• • •
••• ••••• •
••••••• •••••
••••••
••••••
••••••
(a) Crystalline
- •• • • • •
(b) Amorphous (e) Polyerystalline
Figure 4.3 Three types of solids: (a) ordered crystalline and (b) amorpholls materials arc
illustrated by microscopic views of the atoms, whereas (c) a polycryslalline structure is
illustrated by a more macroscopic view of adjacent single-crystalline regions. each of which has
a crystalline structure as in (a).
Ionic crystals
Ionic crystals arc made up of positive and negative ions. The ionic bond results primarily
from attractive mutual electrostatic interaction ofneighboring ions with opposite charges.
There is also a repulsive interaction with other neighbors of the same charge. Attraction
and repulsion together result in a balancing of forces that leads 10 the atoms being in
stable equilibrium positions in such an ionic crystal. As for the electronic configuration
in a crystal, it corresponds to a closed (completely filled) outer electronic shell. A good
example of an ionjc crystal is Nael (salt). Neutral sodium, Na, and chlorine, CI, atoms
have the configurations Nail (ls12s12p63s 1) and Cl l7 (ls12sz2p63s13ps), respectively.
That is, the Na atom has only one valence electron, while one electron is necessary to
complete the shell in the CI atom. It turns out that the stable electronic configuration
develops when the Na atom gives one valence electron to the CI atom. Both of them
become ions, with opposite charges, and the pair has the closed outer-shell configuration
(like inert gases such as helium, He, and neon, Ne). The inner shells are, of course,
completely filled both before and after binding of the two atoms. In general, for all
elements with almost closed shells, there is a tendency 10 form ionic bonds and ionic
crystals. These crystals arc usually dielectrics (insulators).
Covalent crystals
Covalent bonding is typical for atoms with a low level offilling up of the ollter shell. An
excellent example is provided by a Si crystal. As we discussed in the previous chapter, the
electron configunttion ofSi can be represented as core +3s 2 3 p2. To complete the outer
3s z3 pz shell, a silicon atol11 in a crystal forms/our bonds with four neighboring silicon
atoms. The symmetry of the hybrid sp3 orbitals dictates that these neighboring atoms
should be situated in the corners of a tetrahedron as shown in Fig. 4.4. Then, the cenlntl
,,
,
FIgure 4.4 Four Sp3 -hybrid bonding orbitals in a crystal orSi.
Si at0111 Jnd each of its nearest-neighbor Si atoms share two electrons. This provides 50-
called covalelll (chemical) bonds (symmetric combinations ofille sp3 orbitals) in the Si
crystal. The four bonding Sp3 orbitals form an energy band that is completely filled by the
valence electrons. This band is called the valence band. The antisymmctric combination
of the Sp3 orbitals with destructive interference, as in Fig. 3.17. leads to antibonding
states and to the first empty energy band. The latter band is called the condue/ioll band
of the Si crystal. Not surprisingly, a covalent bond of this type plays a major role for Gc,
which is also from group IV of the Periodic Table of elements.
In fact, both of the types of atomic bonding discussed previously may exist simultane-
ously in a crystal, for examplc, in III-V compounds. Indeed, the electron configurations
in Ga and As are core +4s 2 4 p l (Ga) and core +4s 2 4 p3 (As). When a GaAs crystal is
formcd, the As atom gives one valence electron to lhe Ga atom, which makes them both
ions. The Coulomb interaction of these ions contributes to the ionic bonding in III-V
compounds. But now each ion has only two 2s electrons and two 2p electrons, which are
not enough to fill the shell completely. Therefore, thc rest ofthe bonding goes through the
formation of the sp3 .hybridized orbitals. This is the covalent contribution to the crystal
bonding. We can conclude that the III-V compounds are materials with mixed bonding-
partly ionic and partly covalent.
Metals
Metals, such as Na, K, and Ca, consist of ions regularly situated in space. Each atom
contributes an electron to an electron "sea," in which the ions are embedded. The systcm
as a whole is neutral and stable. TI1C clectrons contribute significantly to the binding
energy of metals.
4.3 Crystal lattices 71
Table 4.2 Binding energies for different types of crystals
Type of crystal coupling Crystal Energy per atom (eV)
Ionic Nael 7.9

LiF 10.4
Covalent Diamond, C 7.4
Si 3.7
Gc 3.7
Metallic N. 1.1
Fe 4.1
AI 2.4
Molecular and CH~ 0.1
incrt·gas crystals A< 0.8
There exist other bonding mechanisms in solids, in molecular crystals and inert-gas
crystals. However, these crystals have small binding energies, that is, they arc not stable
enough to be lIsed widely in electronics. In Table 4.2, for comparison, we present binding
energies per atom for different types of crystals. We can see that ionic and covalent
crystals typically have binding energies in the range 1-10 e'Y, whereas molecular and
inert-gas crystals are weakly coupled systems. Metals have intermediate coupling. The
binding energy of a crystal is an important parameter, since it determines the stability of
the crystal, its aging time. the applicability of various treatment processes, etc.
Crystal lattices
Now, we return to the discussion ofcrystal periodicity. The periodic arrangement ofatoms
(ions) in a crystal forms the lattice. The positions of atoms in the lattice are defined as
the sites. In principle, atoms always perform small-amplitude oscillations around the
sites. However, in many cases we can neglect these small-amplitude oscillations and
think of a crystal as a system of regularly distributed atoms (ions). In such a perfect and
periodic crystal lattice, we can identify a region called a unit celJ. Such a unit cell is a
representative of the entire lattice, since Ihe crystal can be built by regular repeats in
space of this element. The smallest unit cell is called the primitive cell of the lattice. The
importance of the unit cell lies in the fact that by studying this representative element one
can analyze a number of properties of the entire cryslal. The primitive cell determines
the fundamental characteristics of the crystal, including the basic electronic properties.
One of the most important properties of a perfect crystalline lattice is its rmns/aliona/
symmelly. Translational symmetry is the property of the crystal's being "carried" into
itself under parallel translation in certain directions and for certain distances. For any
three-dimcnsionallattice it is possible to define three fundamental noncoplanar primirive
translarion vectors (basis veclors) at. a2, and 3J, such that the position of any lattice site
can be defined by the vector R = n I a1 + fl2a2 + fl3a3, where lit. 112, and 113 are arbitrary
integers. If we construct the parallelepiped llsing the basis vectors, 31.32. and 3J. we
obtain just the primitive cell.
/
~-'O~-O,--( f-----~:r_
a
~~
0 0
G-
a\
0 0 0
A
0 0 0 0
a\
Figure 4.5 A two-dimensionallauice. Three unit cells are illustrated by A, Bt and C. Two basis
vectors arc illustrated by al and a2.
/ ('.1 :<0:::,//
. , ,
a 3 ..' '""
,
", " ()
a '" ".
:." '
/ .."-- ::":'~:,'..
' ' '/' " .... </'"
'. "'.
,
"
0'-----'-'
(a) (b) (c)
Figure 4.6 Three types of cubic lattice: (n) simple. (b) body-centered and (e) race-centered.
Basis vectors of the three different primitive cells arc Itl. Ill, and It}.
Translational symmetry is illustrated in Fig. 4.5, where for simplicity we present a

t\'1o-dimensionallanicc of which several unit cells (A, B and C) are shown. The cell A
t
obviously has the smallest magnitudes of vectors 31 and 31. Thus, the basis vectors of
the primitive cell are the vectors 31 and 32 from the unit cell A. An arbitrary lattice site
is at the point R = "Ial + 11232 with integers 111 and "2. To visualize the translational
symmetry of this lattice one can start from any point of the lattice and find all other
equivalent positions in space by just applying translations that are integer multiples of
the basis vectors.
Since many of the crystals used in electronics are of the so~called cubic symmetry.
here we consider briefly such cubic lattices. For them, the unit cell may be selected in the
form of a cube. There are three different types of cubic lattices. The simple cubic lattice
has atoms located at each corner of tile cube, as shown in Fig. 4.6(a). The body-centered
cubic lattice has an additional atom at the center of the cube, as shown in Fig. 4.6(b).
The third type is the face~cel1le,.ed cubic lattice, which has atoms at the corners and at
the centers of the six faces as depicted in Fig. 4.6(c).
Specifically, the basic lattice structure for diamond, C, silicon, Si, and germanium,
Ge, is the so-called diamond lauice. The diamond lattice consists of two face~centered
cubic stmctures with the second structure being shifted by a quarter ora diagonal of tile
4.4 Electron energy bands 73
Figure 4.7 The diamond !auice is of the face-centered type of cubic lattices. The tetrahedral
bonding arrangement of neighboring atoms is clear.
first cube, or by a distance 31/4 + 32/4 + 33/4 from the first face-centered structure;
here vectors a" 3lt and 3J are vectors of the first cube as defined in Fig. 4.6(a). In
Fig. 4.7 the atoms orthe first face-centered structure are shown in black and some atoms
of the second one are gray. Thus, a diamond lattice contains twice as many atoms per
unit volume as does a face-centered cubic lattice. The four nearest-neighbor atoms to
each atom are shown in complementary shading for casier visualization. The parameter
a which characterizes the cubic lattice is the so-called !of/ice cOl/sIal/I. The lattice of
volume V = a 3 consists of eight atoms.
Besides the translations, the crystal symmetry contains other symmetry elements, for
example, specific rotations around high-symmetry axes. In cubic crystals, axes directed
along the basis vectors are equivalent and they are the symmetry axes. It is convenient
to use a system of coordinates built on the basis vectors. Three symmetry axes may be
denoted as (100], [0 I0], and (00 I]. (Here we have used notations that are common in
crystallography: (100] is a unit vector in the x direction, (0 I0] in the y direction, and (00 I]
in the =direction. All important symmetry directions of a cube are shown in Fig. 4.8(a).
The directions of the type (110] and [Ill l are also important crystal directions. It is
evident that these directions arc equivalent to the opposite ones, (ii 0] and (ii il, as well
as to analogous ones. In Fig. 4.8(a), we show the symmetry directions of a cubic crystal.
If the crystal is carried into itself on rotation through an angle 2rr / 1/ about some axis that
passes through the crystal, then this axis is said to be an ,,-fold axis Cn. For example, in
a cubic crystal there are three four-fold axes C4 and four three-fold axes C3 , as shown in
Fig. 4.8(b). The symmetry elements of the lattice make analysis of crystals' properties
much simpler.
4.4 Electron energy bands
Summarizing the above analysis, we conclude that a crystal consists of nuclei and elec-
trons. The valence electrons are collectivized by all nuclei and we can expect them to
Figure 4.8 (a) Symmetry directions for cubic crystals: (b) rOiational symmetry directions in
cubic crystals.
be relatively weakly coupled to atoms. This allows one to think about a crystal as a
system with two relatively independent subsystems: the alom (ion) subsystem and the
electron subsystem. In this section, we will consider the electron subsystem of crystals.
Actually, in a crystal an electron moves in the electrostatic potential created by posi-
tivcly charged ions and all other electrons. This potcntial is frequently referred to as the
clyswllille pOlential, W(r). We will consider one-particle states for elcctrons in an ideal
crystal and we will present a classification oflhese states and find a gcneral form for the
wave functions and cncrgies.
An electron in a crystalline potential

For an ideal crystal the crystalline potcntial is pcriodic with the period of the crystalline
lattice. Lei 3jo with j = I. 2. and3. be thc three basis vectors of the lattice that define the
three primitive translations. The periodicity of thc crystalline potenlial, W(r), implies
that
w(r+ tn;a;)
;:1
= IV(r). (4.1 )
where r is an arbitrary point of the crystal and n) arc some integers. The one-particle
wave function should satisfy the time-independent SchrOdinger equation
.
1t1/t(r)
(til,
= -2mo
- '1- + IV(r) ) 1/t(r) = E",(r). (4.2)
-
where lito is the free-electron mass and 1t is the crystalline Hamiltonian. Equation (4.1)
neglects interactions between electrons and that is why the wave function, t/!(r). is often
called the olle-particle wavefimctioll.
Because of the potential periodicity of Eq. (4.1), the wavcfunctions y,(r) may be
classified and presented in a special form. To find this form we introduce the translation
•
operator. 7d. that acts on the coordinate vector r as
)
.
7dr = r + d. d = L"Ja}- (4.3)
)=1
On applying this operator 10 the wavefunction we find that the function 7dt(r) =
y,(r + d) is also a solution of Eq. (4.2) for the same energy E. Let us assume that
the electron state with energy E is not degenerate. Then, we conclude that the two
wavefunctions 1fr(r) and 1fr(r + d) can differ only by some multiplier, Cd:
",(r + d) = Cd "'(r). (4.4)

From the normalization condition
f I",(r + d)I' dr = I (4.5)
we obtain ICdl! = I. Two different translations dl and d z should lead to the same result
as the single translation d = d l + d 2 , i.e.. C dl Cd: = Cdl~dl' From this result it follows
Ihal Cd may be reprcscmed in an exponential form:
Cd = e'kd = eXP(ik tlljaj). (4.6)

I~I
where k is a constant vector. Thus, from Eq. (4.4) we gCI the wave function in Bloch/orm,
(4.7)
where
(4.8)
One can check that the so-called Blochfimcliol1 uk(r) is a periodic function:
)
lik(r + d') = uk(r). d' = L II jaj.

I-I
Therefore, the stationary one-particle wave function in a crystalline potential has the form
of a plane wave modulated by the Bloch function with the lattice periodicity. The vector
k is called the wavel'ector of Ihe electron in the crystal. This wavcvcctor is one of the
quantum numbers of electron states in crystals.
By applying the so-called cyclic boulldary conditions to the crystal with a number of
periods NJ along the direction a}.
(4.9)
we find for k
(4.10)
These allowed quasi-continuum values of k form the so-eallcdfirsl Brillouin =one of

the crystal. They arc just those energy bands which we discussed above by using simple
qualitative considerations. It is important that the symmetry of the Brillouin zone in

k-space is determined by the crystal symmetry.
Let the one-particle energy corresponding to the wavevector k be £ = E(k). If the
wavevector changes within the Brillouin zone, one gets a continuum energy band; i.c.,
an eleclroll energy band. At fixed k, the Schrodinger equation (4.2) has a number of
solutions in the Bloch form:
(4.11)
where a enumerates these solutions and, thus, the energy bands. Owing to the crys-
tal's periodicity, the Bloch function can be calculated within a single primitive cell.
In Eq. (4.11) we normalize the wavefunction Wed. for the crystal volume V = NVo;
N = N l X N2 X N3 and Vo arc the number and volume of the primitive crystal cell,
respectively. From the normalization of the wavefunction ifla.k(r) one obtains
(4.12)
where the integml is calculated overthe primitive cell. This formula allows one to estimate
the order of the value of lla.k: IUa.kl ~ I.
Thus, through this analysis we have established an extremely important property of
the electrons in crystalline solids: despite the interaction of an electron with atoms and
other electrons, in a perfect lattice the electron behaves much like a free electron. The
electron can be characterized by a wavevector k and, thus, it possesses the momentum ',k.
By considering phenomena that have spatial scales much greater than distances between
atoms (ions) in the primitive cell, we may omit the Bloch function l/a.k and describe the
electrons by a wavefunction in the form of the plane wave iflk(r) = A exp(ikr),just as for
a free particle. However, the wavevector changes inside the Brillouin zone in a manner
that is specific for a given crystal and, in general, the energy dispersions £ = £a(k) can
differ from that of the free electron considerably.
The holes
According to the discussion given in Section 4.2, some ofthe energy bands are completely
filled, while the others are almost empty. For our purposes, two of the bands are of great
importance: the upper filled band and the lowest empty band. They are called the valence,
Ev(k), and the COl/dIlC/ioll, Ec(k), band, respectively.
One of the ways to get an electron into the conduction band is to transfer an electron
from the valence band to the conduction band. Thus, for analysis of the valence band, it
is useful to adopt the concept of (l hole as a new quasiparticle; i.e., by a hole we refer
to an electron missing from the valence band. These quasiparticlcs can be introduced
and described on the basis of simple considerations. If the valence band is full, the total
wavevector of all electrons in the valence band is zero:
k,. = L k; = O. (4.13)
i
where the sum accounts for all occupied valence states. Let us assume that one of the
electrons with wavevector lie is removed from the valence band. The total wavevector of
the valence electrons becomes
k" = Lki = -I<". (4.14)
/
On the other hand, removing this electron is identical to the creation of a hole in
the valence band. One can attribute the wavevector of Eq. (4.14) 10 this hole: k ll = -lie.
Then, the energy ofthe valence electrons decreases by the term E\,(~), and, thus, one can
also attribute the energy EII(kll) = - Ey(~) to this hole. If the energy band is symmetric;
i.e., E\,(k) = E y ( -k), we can write for the hole cnergy
£,(k,) = -£,(1<,,) = -£,.(-1<,,) = -£"(k,). (4.15)
Thus, we can characterize the hole by a wavevector, kll' and an energy, EII(kll), and
consider thc hole as a new quasiparticle created when the electron is removed from
the valence band. In the conduction band, the electron energy, Ec(k), increases as the
wavevector, k, increases. Conversely, in a valence band, near the maximum energy of
the band the electron energy, E\,(k), decreases as k increases. However, according to
Eq. (·U 5), the hole energy increases with the hole wavevector, kn. That is, the hole
behaves as a usual particle. Thus, one can introduce the velocity of the hole, VII =
aEII(kn)/akh, and then employ Newton·s laws, etc. The absence ofa negative charge in the
valence band brought about when all electron is removed makes it possible to characterize
a hole by a positive elementary charge; that is, the holes carry positive electrical charge.
It is worth emphasizing that the similarity between the electrons and holes is 110t
complete: the holes exist as qllasiparticles only in a crystal, whereas the electrons exist
also in other physical media, as well as in the vacuum.
Symmetry of crystals and properties of electron spectra

Usually, the energy dispersion relations, EO'(k), are very complex and can be obtained
only numerically in the context of approximate methods.
Fortunately, the Brillouin zone possesses a symmetry which directly reflects the sym-
metry of the unit cell of the crystal in coordinate space. Ifa crystal is mapped into itself
due to transformations in the form of certain rotations around the crystalline axes and
of mirror reflections, one can speak about the point symmelly of directions in the crys-
tal. In the Brillouin zone, Ihis symmetry generates several points with high symmetry
with respect 10 the transformations of the zone in k-space. The extrema of the energy
dispersion Eu(k) always coincide with these high-sylllmetry points. In particular, this
fact allows one to simplify and solve the problem of obtaining electron spectra. Near
eXlrema, the energy spectnl can be approximated by expansion of E(l'(k) in series with
respect to deviations from the symmetry points. Such an expansion can be characterized
by several constants which define the reciprocal effective Illass tensor:
-I -I -I
"'x., 11I.,y ",x:
(_1_) _
",' ,'J'
III;:; 111;) In y }
-I
(4.16)
"':x "'=y-1 ,,'-.~I
where i and j denote X-, y-, and x-coordinates.

Table 4.3 Symmetry points in group IV semiconductors and III-V compounds
Symmetry Position of extremum in

point k-spacc Degeneracy
r o 1
L ±(~/a)[1111· ±(~/a)[ilil. ±(~/a)[ljll. ±(~/a)[lljJ 4
±y(lnla)[100].±y(2~/a)[010].±y(2~/a)[001]. Iyl < 1 6
x ±(Inla)[lOO]. ±(2~ M[OIO]. ±(Inla)[ool J J
Py
Flgure4.9 The first Brillouin zone of cubic crystals. Points of high symmetry, r, L, and X points.
are shown.
Thus, due to the crystal symmetry, the problem of finding an electron energy spec-
trum E,;r(k) is reduced to the following steps: (a) determination of the high-symmetry
k points of the Brillouin zone; (b) calculation of energy positions of extrema; and
(c) analysis of effective masses or other parameters of an expansion of E(I'(k) within
extrema.
The structure and symmetry of the Brillouin zone for cubic crystals of group IV
semiconductors and of IU-V compounds arc very similar. Figure 4.9 shows the Brillouin
zone of these semiconductor materials. The symmetry points are shown in Fig. 4.9 and
presented in Table 4.3. Evidently, as a result of crystal symmetry. several points have
the same symmetry; indeed, they are mapped into themselves under proper symmetry
transformations. Such a degeneracy of the symmetry points is indicated in Table 4.3. In
particular, the r, L, X, and 1:1 points are of central importance. They give the positions
of the extrema of the electron energy in III-V compounds. Ge, and Si.
The bandstructures for GaAs and Si are presented in Fig. 4.10. The energy dispersions
along two symmetric directions of the wavevcctors (Ill] (from r to L) and (100) (from
r to X) are shown. In each case, the energy, E. is taken to be zero at the top of the
highest valence band, which is located at the r point for both of these materials. Since
Si GaAs
3
2
>"
~
I
£g
.~ • • W."
-----}:
'"
L [' x L [' x
Figure4,10 Bandstruclurcs ofSi and GaAs: the bandgap ofSi is £. = 1.12 eV; and the bandgap
ofGaAs is £. = 1.42 cv.
electrons lend 10 be near the energy minima, one can think of electrons as being located
inside marked regions ark-space. Frequently, these regions in k-space are referred to as
energy valleys, or simply valleys. Materials with several valleys are called many.vofJey
semiconductors. For III-V compounds there is only one valley around the point k = 0
(the r valley); however, in the case of Si there exist six fj. valleys in accordance with
the degeneracy of the 6. points. It is worth emphasizing that also other symmetry points
can playa considerable role in processes occurring far from equilibrium. The highest
valence band and the lowest conduction band are separated by energy bane/gaps, Eg• Let
us postpone the analysis of valence bands for a while and focus on conduction bands.
Conduction bands have different structures for the groups of materials considered in
this section. The main difference is that for Si and Ge the lowest minima are located
at l:!. and L points, respectively, whereas for most III-V compounds there is only one
lowest minimum, which is at the r point. The difference is not simply quantitative;
indeed, it is qualitative and leads to a series of important consequences in the behavior of
electrons.
As discussed previously, within the r, l:!., X, and L points the electron dispersion
curves, Ea(k), can be expanded in series with respect to deviations from the minima,
k~:
Ea(k) = E(k~) +2 1("') -.

m fJ ..
'J
(k, - k,.,)(k j - k j .,). (4.17)
For the r point, we have the simplest c·ase of the so-called isotropic effective mass, m·:
-I ) - -II i .. (4.18)
( ",~. i) - m· IJ'
For l:!., X, and L points, i.e., for l:!., X. and L valleys, lhe reciprocal effective-mass
tensor has only two independent components, corresponding to the longitudinal, fill, and
transverse, /1/1. masses. For example, in the case of Si, for a coordinate system with the
Table 4.4 Energy-band parameters for Si and Ge
Group IV Si Gc
Type of bandgap Indirect Indirect

Lowest minima t1 points L POilUS
Degeneracy 6 4
£, (eV) 1.12 0.664
Electrons
ml/mo 0.98 1.64
mdmo 0.19 0.082
Holes
IIIhh/mo 0.50 0.44
mlh/ 1II 0 0.16 0.28
t1 so (eV) 0.044 0.29
z-axis along the axis where t.he fj. valley under consideration is located and the two other
axes perpendicular to the first, one can obtain
m- l , o o
o ,
",-I o (4.19)
o o ",~l
I
where m I and "'\ are called the IOllgitudinal and transverse effective masses, respectively.
For group IV semiconductors, Si and Ge, the energy parameters are presented in Table 4.4.
The degeneracy, given in Table 4.4. indicates the existence of six and four equivalent
energy valleys for Si and Ge, respectively. According to Fig. 4.9, the L points arc located
at the edges of the Brillouin zone, i.e., only half of each energy valley lies inside the first
Brillouin zone. This reduces the effective number of the valleys to four. In Table 4.4 the
bandgaps, Eg , and the electron longitudinal, "'I, and transverse, ",\. effective masses are
also presented.
For III-V compounds, different situations occur for different materials: some of these
compounds are direct-bandgap crystals and others are indirect materials. Thus, for the
conduction bands ofthese materials the conduction band edge can be found at the r point,
at the L point, or at the X point. In Table 4.5, energy-band parameters for three sets of
II I-V compounds are shown; the direct or indirect nature of the crystals is indicated.
Now we shall consider valence bands. Their structure is more complicated. For group
IV semiconductors and for IU-V compounds, the top of the valence bands at k = 0
has high degeneracy, i.e., several valence bands have the same energy at this point. The
degeneracy occurs because these bands originate frOI11 the bonding of three p orbitals of
the atoms composing the crystals. Thus, if one neglects the interaction between the spin
of the electrons and their motion (so-called spill-orbit interaction), one obtains three
degenerate valence bands, each of which is also doubly degenerate as a result of electron
spin. In fact, the spin-crbit interaction causes a splitting of these six-fold degenerate
states. At k = 0 they are split into (i) a quadruplet of states with degeneracy equal to 4
and (ii) a doublet of states with degeneracy equal to 2. This splitting of valence bands,
Table 4.5 Energy-band parameters for III-V compounds in the As family
InAs GoA, AlAs
Type orgap Direct Direct Indirect

Lowest minima r point r point X points
E, (oV) 0.354 1.42 2.16
Electrons
/II" /1110 0.025 0.067 0.124
Holes
IIIhhl fflO 0.41 0.50 050
11111.1 1110 0.26 0.07 0.26
.6.§<) (cV) 0.38 0.34 0.28
.6. s0 • at k = 0 is shown in Fig. 4.10. One often refers to the lower valence band as the
split-Off valence band. At finite k, the spin-orbit interaction leads to further splitting of
the upper valence band into two branches: the heavy- and light-hole baflds. Parameters
characterizing these bands are the heavy-hole, I1Ihh, and light-hole, JIIlh, masses. The
effective masses of the light hole, 11I1h, and heavy hole, "'hh, as well as the distance to the
split-off band, .6. so • for Si, Ge, and some IIJ-V compounds are presented in Tables 4.4
and 4.5.
It is worth mentioning that, despite the relative complexity of the picture presented
for energy bands, Ihe description of electron properties in terms of the electron and hole
quasiparticles by using several EO'(k) dependences is incommensurably simpler Ihan
operating with the enormous number of valence electrons in the crystal (_10 23 cm- 3).
Direct-bandgap and indirect-bandgap semiconductors

One of the important conclusions, which can be drawn from the energy-band picture
described above is related to the optical properties of the crystals.
In Section 2.4, following M. Planck and A. Einstein, we found that electromagnetic
radiation (light) can be thought of as a flux ofphotons with certain energy and momentum
values given by Eqs. (2.39) and (2.40). The absorption and emission oflighl can be inter·
preted as the absorption and emission of discrele "pori ions" of the light or photons with
a specific energy and momentum. Let us apply these findings to consider light-erystal
interactions. In a crystal absorbing or emitting a photon of sufficient energy, an electron
may be tf'Jnsferred between the valence and conduction bands. For a given frequency of
light, w, such a transition is possible if the energy and momentum conversation laws are
satisfied
E,(k,) - E,(k,) = hw.

k l -k2 =±q.
where k] and k2 are the wavevectors ofthe electrons participating in phototransition; here.
q is the photon wavevector. The sign, + (-), in the second equation stands for photon
emission (absorption). Throughout the whole optical spectral region (from infrared to
ultraviolet light), wavelengths). are much greater than electron de Broglie wavelengths,
as estimated in Chapters 2 and 3. The photon wavevcctors q = 2Jr/)., in turn, arc much
smaller than the electron wavevcctors (Ikll. Ik 2 1» Iql). This property reduces the above
equations to
and E,(k) - E.(k) = hw.
In other words, under absorption and emission of the light electrons transferred between
the valence and conduction bands practically preserve their wavevectors. In other words,
the electron wavevector changes very little. In an energy scheme like that presented
in Fig. 4.10, the processes or light-crystal interaction can be interpreted as vertical
electron interband transitions. Another conclusion following from this analysis is that
the interband phototransitions are possible only for light with energy quanta exceeding
the bandgap, tloo 2: E g. This finding implies that a pure semiconductorcryslal is optically
transparent for light with tioo < £g (). > 2lTcfl/ Eg • with c being the velocity of light ill
vacuum). For some optoelectronic applications, the spectral range near the onset of
absorpt ion/emission of light is critically important.
On combining this analysis with the previously described bandslructures of different
materials we can see that phototransitions induced by light with energy near the bandgap
are possible in semiconductors for which the conduction and valence bands have a
minimum and a maximum, respectively, at the same r point. For example, in GaAs one
can transfer an electron from the valence band to the conduction band directly without a
change in its momentum. Crystals of this type are calleddirect.bamlgap semiconductors.
In contrast, in order to move an electron from the valence band to the conduction band
in Si and Ge, one needs not only to add an amount of energy - greater than the minimum
energy difference between the conduction and valence bands - to excite an electron, but
also to change its momentum by a large amount (comparable to the scale of the Brillouin
zone). Such a semiconductor is called an indirect-bandgap semicondllclor.
Summarizing, the bandstructure of a semiconductor material determines both elec-
trical and optical properties. Manipulation of electrons using light, i.e., optoelectronic
functions, is easier for a direct-bandgap semiconductor, such as GaAs. In contrast, silicon
and other group IV materials are optically relatively inactive. The situation is changed
in nanoscale Si and SiGe structures. where the momentum conservation law is no longer
rigorously obeyed.
Bandstructures of semiconductor alloys

As emphasized in previous discussions, the energy bandstructure of a particular semi-
conductor determines its electrical and optical properties. For naturally existing semi-
conductor crystals such as monatomic Ge and Si, and binary GaAs, their fixed, and
unalterable, energy bandstruclures restrict their applications. One of the powerful lools
for varying the bandstructurc is based on alloying two or more semiconductor materi-
als. Some alloys exhibit we1l·ordcred crystal structures. Though an alloy always has a
disorder ofthe constitutive atoms, contemporary technology facilitates partial control of
Table 4.6 Bandgaps for III-V alloys
Alloy E, (cV)
AI.. Gal_"As 1.42 + L247x

AI.:rlnl_zAs 0.360 + 2.0 12x + 0.698x 2
GU;rlnl_,.As 0.360 + I.064x
,
Ga.. Ln l_.. Sb 0.172 +O.139x +0.415x-
AI.. GU1_.. Sb 0.726 + 1.129,Y + 0.368.\"2
Atlnl_.:r Sb 0.172 + 1.62 Ix + 0.430x 2
Table 4.7 Effective masses for the alloy A1 xGa I _ xAs
Type orminimum Effective mass. ",:/1110
r point 0.067 + D.OS3.\"

X minima 0.32 - 0.06x
L minima 0.11 + D.03x
Heavy hole O.62+0.14x

Light hole 0.087 + O.063x
this disorder and produces high-quality crystals. The properties of such materials can be
interpreted in terms of nearly ideal periodic crystals.
Consider an alloy consisting of two components: At with a fraction x, and B with t
a fraction I-x. If A and B have similar crystalline lattices, one can expect that the
alloy A.rBI_x has the same crystalline structure, with the lattice constant (Ie given by
a combination of lattice constants of materials At (lA, and B t aBo The simplest linear
combination leads to the equation (Vegard's law)
(4.20)
Then. the symmetry analysis can be extended to these types of aHoys. For SiGe alloys
and III-V compounds, this leads us to the previously discussed symmetry properties
of the energy bands. Since the bandstruchlres are similar, one can characterize certain
parameters of the alloy as functions of the fraction x. This approximation is often called
the virlual-o)Jstal approximation. For example, the bandgap of an alloy can be repre·
sented as E:"OY = Eg(x). Such approximate dependences are given in Table 4.6 for III-V
alloys. They correspond to the bandgaps, E g , at r points.
As the composition of an alloy varies, the internal structure of energy bands changes
significantly. For example, in the case of AI.tGa l_.tAs alloys, the lowest energy minimum
of the r conduction band ofGaAs is replaced by the six X minima of AlAs as the value
of x is increased. Indeed, near Ihe composition x ~ 0.4, the alloy transforms from a
direct- 10 an indirect-bandgap material. The x-dependences of the effective masses for
various electron energy minima as well as for heavy and light holes for AI.rGal_.tAs are
presented in Table 4.7.
Clearly, the established capability of fabricating a variety of high-quality materials

provides an excellent tool for modifying the fundamental properties of materials.
4.5 Semiconductor heterostruclures
Further modification and engineering of material properties is possible with the use of
heterostructures. Heterostructures are structures with two or morc abrupt interfaces al
the boundaries between the regions of different materials. With modern techniques for
growth of materials, it is possible to grow siructures with transition regions between
adjacent materials that have thicknesses of only one or two atomic mono layers.
Band offsets at heterojunctions

Let us consider a junction bct\vcen two different semiconductor materials, which gener-
ates an abrupt change in the energy gap as well as an abrupt change in the conduction-
and valence-band energies. These abnlpt changes resuJt in band-offset steps.
To understand the principal novel features brought about by an abrupt energy change
in the energy bandstructure, we need to deviate from the approach of the previous
sections where, while considering the energy bands of semiconductors, we analyzed
energy structures in terms ofrelativeposifiolls of the bands in each ofthe semiconductors.
In this approach, absolute values of the energies were nol important and only relative
positions of the bands were taken into account. However, if two different materials are
brought together, the absolute values of energies become critically important. There is a
simple way to compare energy bands of different materials. Let us introduce the vacullm
level of the electron energy, which coincides with the energy of an electron "outside"
of a material. It is obvious that the vacuum level may be taken to have the same value
for any material. One can characterize the absolute energy position of the bottom of the
conduction band with respect 10 this level, as shown in Fig. 4.11. The energy distance
between the bottom of the conduction band and the vacuum level, X, is called Ihe electron
affinity. In olher words, the electron affinity is the energy required to remove an electron
from the bottom of the condUClion band to outside of a material, i.e., to the so-called
vocullm level. Thus, if we know the electron affinities for different materials. we know
the values of the conduction-band bottoms with respect to each other.
With this definition of electron affinity, one can calculate the discontinuity in the
conduction band at an abrupt heterojunction of two materials, A and B:
(4.21)
where XA.B are the electron affinities of materials A and B. Similarly, one can calculate
Ihe discontinuity of the valence band for the same helerojunction:
fj.£.. = £\'.0 - £".A = XB - XA + fj.£" (4.22)
where fj. £g = E g.A - £g.8 is the bandgap discontinuity for the heterojunction. with
E g.A and £g.8 being the bandgaps of materials A and B, respectively. Thus, ifthis simple
4.5 Semiconductor heterostructures 85
Vacuum level
cc
""
~
X ~
c
eu
-w
u
Bottom of the conduction band

c
-8 ..
Q.
~
"- "
c
"
Eg :2 ~
or>
5 -
••••••
••••••
••••••••••••••••••
••••••••••••••••••
••••••••••••••••••
Figure 4,11 This diagram illustrates the electron affinity and vacuum level in it crystal.
approach -the eleclroll~{ifJillityrille - is applicable to a pair of semiconductor materials,

one can calculate band offsets for an ideal heterojunction. Furthermore, ifthree materials,
say A, B. and C, obey this rule, the following "transitivity" property is valid:
6E,(A/B) + 6E,(B/C) + 6E,(C/A) = O.
where fiEvCAjB) is the valence-band discontinuity at the AlB interface, Hence, it is

possible to calculate the band offset for one of three junctions if parameters for two of
them are known,
Unfortunately, this rule fails for many semiconductor pairs. One reason for this failure
is the dissimilar chamcter ofchemical bonds in adjacent materials. The formation of new
chemical bonds at such a heterojunction results in charge tmnsfer across this junction
and the consequcnt rcconstruction of encrgy bands, which leads to the breakdown of the
electron-affinity rule. In real hetcrojunctions, band offsets can depend on the quality of
the interface. conditions of growth, etc.
On combining different values of the electron affinity and the energy bandgap, we
can expect different band line~ups at the interface between two semiconductor materials.
In Fig. 4,12, sketches of the three possible types of band discontinuity are presented.
The most common line-up is of the "straddling" type presented in Fig. 4.12(a), with
conduction- and valence~band offsets of opposite signs, and Ihe lowest conduction~band
slates occur in the same part of the structure as lhe highest valence~band states. This case
is referred to as a type I heterostructure. The most widely studied hcterojunctiol1 system.
GaAslAI.\.Gal_xAs, is of this kind for x < 0.4. The next sketch, Fig. 4.12(b), depicts a
heterostructure where the lowest conduction-band minimum occurs on one of the sides
(e)
Rgure4.12 Three types ofinlerfaces: type I (a): type: II (b); and broken-gap lineup (e). The
energy bandgaps orthe materials A and B arc indicated in (b).
-eF~-O
A
+eFh-O
(b)
Figure 4.13 (a) A external electric fIeld. F, simply tilts the conduction and valence bands ora
semiconductor. The forces -eF' and +eF acting on the electron and the hole are equal in
magnitude and opposite in direction. (b) The same direction of forces -eFt and +eFb on the
electron and the hole caused by quasiclcctric field in the conduction band, Fe. and in the valence
band, Fh. (e) Forces -eFe and +eFh ofopposilC directions for electrons and holes.
and the highest valence-band maximum on the other, with an energy separation between
the IwO less than the lower of the two bulk bandgaps. This case represents a type /I
Iteterostmctllre. The combination AIAs/AI.lGal_.lAs for x > 0.4 and some Si/Si., Gc I-.l
structures arc of this kind. Figure 4.12(c) illustrates a broken-gap line-up, in which the
bottom of the conduction band on one side drops below the top of the valence band on
the other. The example of this band Line-up is given by InAs/GaSb. with a break in the
forbidden gap at the interface of the order of 150 meV.
Graded semiconductors
Very often. graded semiconductors are used instead of abrupt heteroimerfaccs. To illus-
trate the idea. consider first a homogeneous piece of a semiconductor, say. a piece of
unifonnly doped silicon, but with an electric field applied. Then. the band diagram looks
like that illustrated in Fig. 4.13(a) and is represented simply as two parallel tilted lines
corresponding to the conduction- and valence-band edges. The separation between the
two lines is the energy bandgap of the semiconductor: the slope of the two band edges
is the elementary charge e multiplied by the electric field F. When an electron or a hole
4.6 lattice-matched and pseudomorphic heterostructures 87
Table 4.8 lattice constants for cubic

semiconductor materials (T = 300 K)
Semiconductor Lauice constant (A)
SiC 3.0806
C 3.5668
Si 5.4309
GaP 5.4495
GaAs 5.6419
Go 5.6461
AlAs 5.6611
InP 5.8687
InAs 6.0584
is placed into this structure, a force -eF acts on the electron and +eF acts on the hole;
the two forces are equal in magnitude and opposite in direction.
Slopes of conduction- and valence-band edges arise in the case of a graded transition
from one material to another. Graded transitions from a narrow-bandgap to a wide-
bandgap semiconductor that correspond 10 the abrupt heterojunctions of Figs. 4.12(a)
and 4.12(b) are shown ill Figs. 4.13(b) and 4. I3(c). As is obvious from Figs. 4.13(b)
and ·U3(c), in the case of graded helerostructures there is a built-in electric field that
acts on electrons and holes. This field is called qUllsielectric. The quasielectric field
does not exist in homogeneous crystals; that is why graded heterostructures can be used
for new devices where the existence of a built-in electric field is required. Examples of
materials used in graded nanostructure devices are Si..-GCI_x and AI.~Gal_.tAs, where
x changes in the direction of growth. Graded structures and the accompanying quasi-
electric forces introduce a new degree of freedom forthe device designer and allow him to
obtain effects that are basically impossible to obtain using onJy external (or real) electric
fields.
4.6 Lattice-matched and pseudomorphic heterostructures
Now we shall consider some of the principal problems that arise in the fabrication of
heterostructures. In general, one can grow any layer on almost any other material. In
practice, however, the interfacial quality of such artificially grown structures can vary
enormously. Even when one fabricates a structure from two materialsofthe same group or
from compounds ofthe same family, the artificially grown materials ofthe heterostructure
may be very different from the corresponding bulk materials. First of all, the quality of
the materials ncar heterointerfaces depends strongly on the ratio of lattice constants for
the two materials.
In Table 4.8. lattice constants for several group IV semiconductors and III-V com-
pound semiconductors are presented; all of the cases presented represent cubic crystals.
The lattice constants for some other materials can be found from Fig. 4.14. Depending on
the structural similarity and lattice constants of the constituenlmaterials, there exist hvo
3.0
2.5 AlP CdS

•
,-.. 2.0
>
~ 1.5
CdTe
5i
kl'"1.0
Go GaSb
0.5
InA.<
5.6 5.8 6.0 6.2 6.4 6.6

ao (A)
FIgure 4.14 Room-temperature bandgaps. E,.. as functions of the lattice constant. ao, for selected
III-V and II-VI compounds and selected group IV materials and their alloys.
essentially different classes of heterointerfaces: lattice-matched and lattice-mismatched

materials. Prior to an analysis of both classes, we highlight other factors affecting the
quality and usefulness ofheterointerfaces.
Valence matching
Since there arc still no rigorous rules for how one can realize a given level of quality for
heterojunctions, we consider a few examples, which illustrate the problems.
If lattice matching were the only obstacle, the Ge/GaAs system would be the ideal
hetcrosystem, because, according 10 Table 4.8, it would allow one to realize the ideal
combination of group IV semiconductors and III-V compounds. Indeed, on the basis
solely oflattice-constanlmatching, the Ge/GaAs system appears to be the most promising
candidate. However, it turns out that there is the problem of chemical compatibility for
this heterostructure. Covalent bonds between Ge on the one hand and Ga or As on the
other arc readily formed. but they arc what could be called valetlce·mismllfched bonds,
meaning that the number of electrons provided by the atoms is not equal to the canonical
number of exactly two electrons per covalent bond. Hence, the bonds themselves are not
electrically neutral. Consider a hypothetical idealized (00 I).oriented interface between
Ge and GaAs. wilh Ge to the left of a "mathematical plane" and GaAs to the right, as
shown in Fig. 4.15. In GaAs, an As atom brings along five electrons (resulting in 5/4
electrons per bond) and is surrounded by four Ga atoms, each of which brings along
three electrons (3/4 electron per bond), adding up to the correct number of two (8/4)
electrons per Ga-As covalent bond. However, when, at a (00 I) interface, an As atom
has two Ge atoms as bonding partners, each Ge atom brings along one electron per
bond. which is half an electron more than is required for bonding. Loosely speaking,
the As atom does not "know" whether it is a constituent of GaAs or a donor in Ge. As
a result, each Ge-As bond acts as a donor with a fractional charge, and each Ge-Ga
L- [OOIJ
eGa OAs
Figure 4.15 Departure from electrical neutrality at a "mathematically planar" (00 I)-oriented
Ge/GaAs interface. The different atomic species. Ga or As atoms. and Ge atoms. do not bring
along the correct number of electrons to form electrically neutml Ga--Ge or As-Ge cowlent
bonds of two electrons per bond. Reprinted with pcnnission from W. A. Harrison, E. Kraut
et al.• "Polar hetcrojunction interfaccs" Phys Rev. B. 18(8),4402 (1978). 1[l1978 by the
American Physical Society.
bond acts as an acccptor with the opposite fractional charge. To be electrically neutral,
a Ge/GaAs interface would have to have equal numbers of both charges, averaged not
only over large distances, but also locally. Given chemical bonding preferences, such an
arrangement will not occur naturally during epitaxial growth. If only one kind of bond
were present, as in Fig. 4.15, the interface charge would support a large electric field
of 4 x 107 V cm- I • Such a huge field would force atomic rearrangements during the
growth, trying to equalize the numbers ofGe-As and Ge-Ga bonds. However, these
rearrangcments will never go to completion, but will leave behind locally fluctuating
residual charges, with deleterious consequences for the electrical properties of materials
and any devicc applications.
Interfaces with perfect bond-charge cancellation may readily be drawn on paper, but
in practice there are always going to remain some local deviations from perfect charge
compensation, leading to performance·degrading random potential fluctuations along
the interface. This argument applies to other interfaces combining semiconductors from
different columns of the Periodic Table of elements.
The above discussion pertains to the most widely used (00 I)-orientcd interface. The
interface charge at a valence-mismatched interface actually depends on the crystallo-
graphic orientation. It has been shown that an ideal (112) interface between group IV
and lll-V compounds exhibits no interface charge. An important example is the GaP-
on-Si interface that has a sufficiently low defect density: as a result, it is used in various
devices grown on Si. After these comments, we return to the discussion of the role and
importance of Ian ice matching.
:>-0 A
(b) (e)
Figure 4.16 Lattice·matched materials (a) and lattice-mismatched materials (b); the resulting
structure is strained (pscudomorphic) if the upper layer B adopts the lattice of substrate A (c).
Lattice-matched materials
For lattice-matched structures, the lattice constants of the constituent materials are nearly
marched; i.e., the lattice constants are within a small fraction of a percent of each other.
There is no problem, in general, in growing high-quality heterostructures with such
lattice-matched pairs of materials. By "high-quality" we mean that the interface struc-
ture is free of lattice imperfections such as interface defects, etc. Such imperfections
result in poor electrical and optical properties and may lead to fast and widespread
degradation of the structure. Figure 4.16(a) illustrates a lattice-matched layer B on a
substrate A. One can expect that the layer can be grown on the substrate ifboth mate-
rials are from the same group and the binding energies and crystal structures are very
similar.
According to the data of Fig. ·U4 and Table 4,8, the AIGaAs/GaAs systcm is an
example of a lattice-matched material. The system has a very small mismatch in the
lattice constants of only about 0.1 % over the entire range of possible AI-to-Ga ratios
in the AIGaAs. As a result, such heterostructures can be grown free of mechanical
strain and significant imperfections, Hcnce, these structures provide a practical way
of tailoring bandstructures. In addition to these tailored electronic parameters, elastic
and other lattice properties can be different in layers composing such a lattice-matchcd
heterostructure.
Lattice-mismatched materials
TI1C case of lattice-mismatched stnlctures is characterized by ajillife latrice mismatch.
Figure 4, 16(b) depicts this case. If one tries to match these lattices, strain in the plane
of growth and a distortion along thc growth axis arise. Thus, one obtains a strained
layer with a lattice deformation. The lattice-mismatched stnlcture can be characterized
also by the relati\'e mismatch of the lattice constants of the substrate, aA, and the
epilaycr, au:
<= • (4.23)
"A
Consider an elastic defornlation of a lattice. It can oc characterized by the vector of
relative displacement u. The displacement defines how any lattice point r moves to a
new position, r1 = r + u, as a result of the deformation. Different regions of the crystal
can be deformed differently. Thus, the displacement depends on coordinates: u = u(r).
In fact, only relative displacements arc important. Thcy are detennined by the straill
tellsor:
IIi) = ~ (:;; + ~~:). (4.24)
where the tensor IIi; has the following components:

lI:u II x .. "x~
Ulj - ")'x U • l' 11.1,:: (4.25)

U:x " :)' II::
=
Here i and j denote x, y. and components; ~i = x. y, and for i = 1.2, and 3, =
respectively. In this discussion, we consider only diagonal components of lI,j. They
determine a change in the crystal volume, from V to V', produced by the strain:
V' - V
8 = -'---,--- (4.26)
V
The clastic energy density (elastic energy per unit volume) of a crystal may also be
expressed in terms of the strain tensor. For cubic crystals, this energy is given by
U= ~ Cll (II;... + ll.~,), + u;:) + C44 (1/;.1' + I/.~: + 1/;.:)

+ Cl2 (UUll yy + Uull:: + II 1)'11::). (4.27)
where Cll. ell, and C44 arc elastic COllstallis or elastic moduli of the crystal. Equa-
tion (4.27) is a generalization of the expression for the potential energy of an elastic
isotropic medium derived in Section 2.3. The elastic constants Cll and C4-1 arc responsi-
ble for anisotropy of crystals. this is why they equal zero for an elastic isotropic medium.
As a result, Eq. (2.17) is recovered in this case with A = Cll.
The stress tellsor is defined in tcrnlS of derivatives of the elastic energy density with
respect to strain-tensor components:
au
alj =-. (4.28)
all,;
Boundary conditions at a surface or at an interface may be formulated in tenns of the
stress tensor:
(4.29)
where N is a vector perpendicular to the surface and f is an external force applied to the
surface.
These equations are sufficient for calculations of the strain of a layer A grown on
a mismatched substrate B. Let the lattice constants of these two materials be aA and
aa. respectively. In this discussion, both materials arc assumed to be cubic crystals and
the direction of growth is along the [001] direction. If the layer A adopts the lattice
periodicity of the substrate B, the in-plane strain of the layer is
U.u = tin'
"B
= fill = I - - . (4.30)
",
ll1ere should be no stress in the direction of growth. Thus, from Eq. (4.29) it follows that
a;; = O. On calculating a;; from Eqs. (4.27» and (4.28), and from the obtained result,
a;; = CIIII;; + Cl2(I1...... + lI yy ), we find the strain in the direction perpendicular to the
layer:
2Cl2
II;; = - - u I I ' (4.31)
CII
Thus, the strain can be found through the mismatch of the lattice constants.
The strain result's in two types of effects: (I) the strain can generate various imperfec-
tions and defects; and (2) the strain in the layer leads to a change in the symmetry of the
crystal lattice, for example, from cubic to tetragonal or to rhombohedral, etc. Of course,
the latler effect can modify the energy bandstrueture of tile layer.
Strained pseudomorphic heterostructures

Here, we consider imperfections generated by strain from a lattice mismatch. In order
to understand the nature of the formation of imperfections in a layered structure, let
us consider the characteristic energies of the structure. First of alL a layer grown on a
substrate with a mismatched lattice should possess extra elastic energy, Eel. caused by
the strain. This energy is a function of the thickness of the layer, d, and increases with
increasing d. In the simplest case of uniform strain, the elastic energy can be calculated
through its density U: Eel = U X d X S, where S is the area of the layer. On the other
hand, the generation of misfit defects requires some energy. Let us denote this energy by
£im' If the extra elastic energy exceeds the energy associated with the imperfection. i.e., if
£el(d) > Elm, the system will relax to a new state with lower energy and imperfections
will be generated. That is, the extra strain energy is the main physical reason for the
instability and degradation of heterostructures fabricated from materials with a large
mismatch of lattice constants.
Since the value of Eim remains finite even in thin layers, for certain thicknesses we
may get Ed(d) < £nn. Thus, there is not sufficient strain energy and imperfections will
not be genernted. Such strained heterostructures can be of high quality. Hence, in some
approximations, for each pair of materials there exists a critical thickness of the lay-
ers, d cr : if d < d cr the lattice mismatch is accommodated by the layer strain without
the generation of defects. The corresponding layered systems are called pselldo11lorphic
heterostruclllres. In general, a pseudomorphic layer of material possesses some charac-
teristics similar to those of the substrate and may possibly have the same lartjce structure
as the substrate material. In our case, a crystalline semiconductor layer grown on another
4.6 lattice·matched and pseudomorphic heterostructures 93
semiconductor takes on the in-plane lattice periodicity of the substrate semiconductor.

Figure 4.16(c) illustrates the case when the deposited layer adopts the lattice periodicity
of a substrate material. Examples of such systcms are the Gal_.rAlxAs/Gal_xln.rAs and
GaAslGal_.rln.rAs structures. In fact, these heterostructmes arc used to improve thc char-
acteristics ofthe so-called hererojullcrion-field-effecr trallSistors which wi II be considered
in Chaptcr 8. In spite of significant mismatches of lattice constants, thcse structures arc
virtually free of interface defects, because of the small, nanometer-scale. thicknesses
of the pseudomorphic layers used in the fabrication of functioning heterojunction-field-
effccttransistors.
It is somctimcs possible to grow defect-free systcms with layer thicknesses exceeding
the critical thickness. However, such systems are metastable and this may lead to device
degradation as a result ofthe generation of misfit defects driven by temperature effects or
other external perturbations. Central to the stability of pseudomorphic structures is the
qucstion of whcther or not the strain energy leads to damage of the materials when the
structurcs arc subjected to various forms ofexternal stress and processing. The experience
accumulated in this field shows that in the case ofsmaU strajn energy the heterostructures
arc stable. For example, in the case of the GaP/GaAsP layered system, the strain energy
is about 10- 3 eV per atom. Since this quantity is rather small in comparison with the
energy required to remove the atom from its lattice site, this system can be stable for
sufficiently thin layers.
The above-discussed strain states are shown in Fig. 4.17 as a function ofx for GexSi l _.r
layers grown on Si substrates. The "phase diagram" - the critical thickness of the layer
versus the Ge fraction - consists of three regions: strained layers with defects at large
thicknesses, nonequilibrium (metastable) strained layers without defects at intermedj-
ate thickncsses, and equilibrium and stable layers without defects at small thicknesses.
According to these results, a stable Ge layer on Si (the largest misfit) can not be grown
•
with a thickness greater than lOA or so.
Let us consider the Si/Ge system in more detail. This system is very interesting
and important because it has opened new horizons for silicon nanotechnology and Si-
based applications. The data of Table 4.8 show that heterostructures based on Si and Ge
materials should be designed so that they arc always pseudomorphic.
First of all, the stability and quality of these Si/Ge pscudomorphic heterostructures
depend strongly on the thicknesses of the straincd layers, as discussed previously. In
fabricating Si/Gc structures, one grows specific numbers ofSi and Ge atomic monolayers.
Thus, layer thicknesses can be characterized by the numbers of these monolayers. Let
fl and III be the numbers of Si and Ge monolayers, respectively. This system is known
as the Sin/Gem super/altice. The second important factor that determines the quality of
these structures is the material of ,he s/lbstrate on which the superlattices are grown.
We have discussed the case ofGe... Sh_.r layers grown upon a Si substrate; see Fig. 4.17.
For the fabrication of SilGe superlattices, the substrates of choice are frequently either
Ge.rSi l _-,· alloys or GaAs.
Let us consider GcxSi I-.r as a substrate. The clastic energy ofa strained system depends
on the alloy composition of the substrate. Figure 4.18 illustrates this dependence for
various numbers of mono layers for the symmetric case, 11 = m. In accordance with the
Misfit (%)
o 1 2 3 4
."t 1 ~1J11
•
~
~
.-"
~
F 1000 A
-
-U
.-
.- 100 A
loA '--~'--~'--'----''--'----''--~
o 0.2 0.4 0.8 0.6 1.0
Germanium Fraction. x
Flgure4.17 The stability-strain diagram fora Ge.rSil_~ layer on Si substrate. After R. People and
1. C. Bean. "Erratum: Calculation of eriticallayer thickness versus lattice mismatch for
Ge... Si 1_.,/Si strained-layer heteroslructurcs" Appl. Phys. Leu. 47. 322 (1985). Appl. Phys. Leu.
49,229 (1986). Reprinted with permission from R. People and 1. C. Bean, Appl. Phys. Len. 49
229 (1986). © 1986 American Institute of Physics.
4
~
"" 3
""
.'"
'"'" 2
1/=1
a 0.5 1
Ge fraction
Figure 4.18 The elastic energy of strained Ge"/Si" superlauices on Gc.,Si 1_.' substrate with
various nwnbers of monolayers n as a function of the Ge fraction in the Ge~Sil_. substrate. The
numbers on the curves indicate II.
4.7 Organic semiconductors 95
previous discussion, the elastic energy increases with increasing thicknesses of strained
layers for a given substrate material. Because of this, one employs superlanices with
:s
a few monolayers: 2 (11.111) :s5. Figure 4.18 also shows a nontrivial dependence of
the strain energy on the alloy composition of the substrate; the minimal strain energy is
expected for x in the range from 0.4 to 0.6.
Another important characteristic ofpseudomorphic Si/Ge structures is the distribution
of the elastic energy over the monolayers of the superlattice. It was shown that the most
homogeneous distribution over layers occurs for the Si/Ge alloy with x ~ 0.5. From
this point of view, Sio.sGeo.s substrates are preferable. However, these results depend
strongly on the oricntation of the substrate material. Oftcn, the direction of growth on
the substrate is chosen to be the (00 I] direction.
Although the technology of the growth and fabrication ofSi/Ge structures is still in a
developing stage, from our short analysis one can see that there is some qualitative and
even quantitative knowledge concerning the behavior responsible for the stability and
perfection of these structures.
lattice-matched heterostructures
Let us return to Fig. 4.14 and discuss lattice-matched hetcrostructures in more detail.
From this figure, we can determine the Ian ice constants of various compounds. First of
all, one can see that the GaAsIAlAs system is really unique because the lattice constants
have almost identical values. In order to achieve lattice matching for other cases, it
is possible either to combine a binary compound and a ternary compound, or to use
ternary-ternary compounds having appropriate ratios of atomic species within each
layer. For example, in the case of GaLnAslinP structures, Ian ice matching is achieved
exactly only for G3{).47LnO.S3As, in which the ratio ofGa to Ln is 47 to S3 in the GalnAs
layer; for the other ratios, the GaInAs layer is not lattice-matched with the InP. Moreover,
the wide-bandgap Gao.SdnO.49P material is compatible with the narrow-bandgap GaAs
material.
In conclusion, the broad range ofpossibilities for controlling bandgapsand band offsets
for both electrons and holes, as well as electron and hole effective masses, provides
the basis for cnergy-band engineering. Through such energy-band engineering, it is
possible to design and fabricate high-quality heterostructures with designated optical
and electrical properties. If one can not achieve the desired properties using lanice-
matched compositions, it is possible to employ strained pseudomorphic structures.
4.7 Organic semiconductors
In recent years, organic molecules have been shown to have properties that make them
suitcd for novel electronic and optoelectronic devices. Such novcl devices include organic
Iight-emitting diodes, electronic circuits in mechanically flexible layers. crystal displays,
novel molecular electronic devices, carbon-nanotube-based devices for data processing,
and bioelectronic devices.
J-I
a-bond
Figure4.19 A schematic diagram ofe21-Lt (cthcne) illustrating (T·bonds and n-bonds between
carbon atoms.
Some organic molecules - including a variety of hydrocarbons - have been shown

to have properties, such as relatively high conductivity and states with energy differ-
ences, 6£ = "W. corresponding 10 near-infrared, visible, and ultraviolet wavelengths,
A = 2rrc/w, that make them suited for novel electronic and optoelectronic devices. These
organic molecules may be deposited and bound on surfaces in specified patterns, llsing
techniques described in Chapter 5, so thaI they form patterned layers with nanoscale
dimensions. ll1Us, active layers in devices may be made very thin so that carrier transit
time through a device is reduced, making possible very fast devices. Many organic poly-
mers are flexible and it is thus possible to make electronic and optoelectronic devices on
flexible thin films. Flexible displays - such as a computer display that may be rolled up
into a pen - represent one of the novel applications of organic molecules in electronics
and optoelectronics.
TIle importance oforganie molecules which form conducting polymers is illustrated by
the fact that Alan Heeger, Alan MacDiarmid, and Hideki Shirakawa won the Nobel Prize
for Chemistry in 2000 for their pioneering research on conducting polymers. Working
together, they learned how to make very pure samples of polyacetylene and found ways
of "doping" them with bromine by exposing them to bromine vapor. They discovered
that it was possible to increase the conductivity of polyacetylene by seven orders of
magnitude! How does this work? To understand this, consider the a- and 7T-bonds in a
simple hydrocarbon such as ethene, C2 H4. as depicted in Fig. 4.19. The localized a-bond
leads to the bonding of the two carbon atoms in a simple hydrocarbon such as ethene.
The extended 7T-bonds are found in Illany hydrocarbons; these molecules are said to be
7T-conjugated. The extended IT-bonds are delocalized in space along the length of the
molecule and arc thus similar to the extended energy bands that arc formed when atoms
arc combined to form a crystaL Thus, these extended 7T-bonds open the way for charge
transport along the polymer chain.
Polyacetylene, shown in Fig. 4.20, is a hydrocarbon that has alternating single and
double bonds leading to slightly di fferent C--C interatomic bond distances and therefore
to energy levels separated in energy by different amounts, 6 E, as discussed in Chapter 3
for bonding and antibonding orbitals; see Fig. 3.17. The differences between the energy
4.7 Organic semiconductors 97
--,,-bonds
Alternating
r1~\--7';!>--'y"'V~\--7'~:"-single and
double bonds
--,,-bonds
Figure 4.20 A polymer with carbon backbone having alternating single and double bonds
between carbon atoms; such n-conjugated polymers have extended wave functions facilitating
charge transpon along the chain. Polyacetylene is a case in point.
Region of 1t-bonds
- - and charge transport
Alternating
",f-"",...,:;:, ~-:-y-vJU",-<;?-::-_ single and
double bonds
Region of 1t-bonds
--and charge transport
Agure 4.21 A n-conjugated polymer (without n-bonds shown) with an acceptor impurity. The
acceptor atom - such as bromine or iodine - attracts and captures one of the electrons in the
extcnded n-statcs and results in a region where there is a deficit of electron charge - a hole. The
captured clcctron is relatively immobile compared with the holc. which movcs along the
polymcr.
levels differ from one hydrocarbon polymer to another, making it possible to produce
conducting polymers with different electrical and optical properties; n-conjugated poly-
mers -like polyacetylcne - have overlapping n-orbitals that conduct charge. The doping
of a n-conjugated polymer with an acceptor impurity is illustrated in Fig. 4.11. The
acceptor removes an electron from the polymer and leaves the polymer with a deficit
of negative charge, i.e., with a positive charge - a hole (more detailed discussion about
doping is given in Section 5.3). This positive charge aUracts neighboring electrons and
produces a local distortion of charge near the positive charge. This distorted charge
density is known as a po/aro". A polaron propagates along the molecule in a manner
that bea.rs some similarity to the propagation of a hole in a semiconductor. Indeed, the
polaron behaves as a positive (screened) particle with an effective mass that is different
from the mass of the electron. It is observed for polymers like polyacetylene that the
mobility of the hole is high and resulting conductivities are close to those of metals.
Thus, doping leads to conductivities in n-conjugated polymers thai make them useful in
electronic devices.
4.8 Carbon nanomaterials: nanotubes and fullerenes
In this section we will study new nanoscale objects - carbon-based nanomaterials. These
include the so·called carbollllal1olllbes 3ndji,lIerenes. These !lanomaterials exhibit many
unusual properties. In particular, the carbon nanotubes exhibit metallic, semiconductor,
and dielectric properties and have great promise for applications in nanoeleclronics.
We start with a brief review on carbon and ordinary carbon materials. Carbon, C, is
a group IV element. as is silicon, Si. Carbon is known to be associated with rich and
diverse chemistry and carbon atoms participate in the formation of a great number of
molecules. Befofe the discovery of carbon-based nanomatcrials, there were only twu
known forms of solid carbon: diamond and graphite. Diamond is a dielectric with a
very large electron energy bandgap (......6 eV). Diamond can be p-doped, thus it should be
considered as a wide-bandgap semiconductor. It is an indirect-bandgap semiconductor.
In contrast, graphite is a semi metal. The structural and electronic properties of both
of these carbon materials are not very promising for common electronic applications.
However, the wide-bandgap properties of semiconducting diamond can be exploited in
high-temperature electronics.
Carbon nanolubes
Actually, carbon nanotubes have the form of seamlessly rolled single sheets of carbon
atol11s. The cylindrical sheets may have diameters of only a fC\.... nanometers. They arc
very small objects that exhibit many different structures and properties. These nan-
otubes have no macroscopic analogues, such as the bulk semiconductors that served as
analogues for traditional semiconductor nanostructures. Thus, to study the stnlcture of
the carbon nanotubes, one should use the most advanced microscopy techniques: atomic
force microscopy and sClllllling tunneling microscopy, which we will consider in the next
chapter.
A carbon nanotllbe is composed ofcarbon atoms. A defect-free single-walled nanotube
consists of a single cylinder and is characterized by the tube diameter and a quantity
known as thehelicity. Tounderstand the structure ofcarbon nanotubes, we wilJ start with a
two-dimensional graphite sheet, shown in Fig.4.22(a). The single sheet ofgraphite, called
graphene. has the form of a honeycomb-like lauice. Let 31 and 32 be the graphene lattice
vectors and /I and m be integers. The diameter and hclicity oflbe nanotube arc uniquely
characterized by the vector C = 1131 + m32 := (n. m) that connects cfystallographically
equivalent sites on a two-dimensional graphene sheet. Here 31.2 are in units of (to J3 with
(to being the carbon-carboll distance. By using the vector C a carbon nanotube can be
constructed by wrapping a graphene sheet in such a way that two equivalent sites of the
hexagonal lattice coincide. The wrapping vector C defines the relative location of these
two sites. In Fig. 4.22(a), the wrapping vector C connects the origin (0, 0) and the point
with coordinates (11,7). Thus, a nanotube denoted by indices (11,7) is formed. A tube is
called an armcllllirtube if n equals m, and a zigzag tube in the case of 111 = O. Wrapping
vectors along the dotted lines leads to tubes that arc of zigzag or armchair fOfm. All other
wrapping angles lead to chiral tubes whose wrapping angle is specified relative either to
4.8 Carbon nanomaterials 99
T lip II
zigll1g
,
•
(O.7)
,
(a) anndtalf (b) zig:Lllg
Figure 4.22 (a) The relation between the hexagonal carbon Ian ice and the chirality of carbon
nanotubes. A carbon nanotubc can be constnlcted from a single graphite sheet, called graphene,
by rolling up the sheet along the wrapping vector C. (b) Fragments or"amlchaii' and "zigzag"
carbon nanotubcs.
the zigzag direction (} or to the armchair direction t{! = 30° - (7. Both (} and the wrapping
angle (cltirtl! angle) Vt arc in the range (0. 30°) as a result of the hexagonal character
of the carbon two·dimcnsional lattice of the graphene. Dashed lines arc perpendicular
to C and run in the direction of the tube axis indicated by the vector T. The solid
vector H is perpendicular to the armchair direction and specifies the direction of nearest-
neighbor hexagon TOWS. The angle between T and 11 is the chiral angle l/J. The unit
cell of a nanotube can be constructed by finding the smallest lattice vector T which
connects equivalent points of the lattice. It can be shown that this vector is given by
T = [(II + 2/1/ )a, - (211 + //I )a,l/q. (4.32)
where a 1 and 32 are the basis vectors of the graphene lattice; q is defined as follows:
q =
I'·3/,
if n - m is not a multiple of 3/.
if" - III is a multiple of 3/.
where / is the greatest common divisor of 1/ and m. The number of hexagons, N, in the
unit cell is
2
N = 2(n + m 2 + "m). (433)
q
The tube radius, R, is given by
R = _IC_I = ..:,J3,...3a-"o,/'II""-+-'-11""""+-'-"-11 . (4.34)

2rr 2rr
The length of tile carbon-carbon bond can be estimated as ao ~ 0.14 nm. It is easy to see
that, for a general chiral nanotube, the number of atoms contained in the unit cell is very
large. For example, an (8, 7) tube has a radius of about 0.57nm and contains 676 atoms
in the unit cell. In determining the number of atoms, we have taken into account that
each hexagon contains nvo carbon atoms: II = 8,111 = 7." - //I = I, q = I, N = 338.
and the number of atoms is 2N = 676. In Fig. 4.22(b), the structures of two particular
examples of the armchair and zigzag types ofnanonlbes are sketched.
Agure 4.23 Single-walled carbon nanotubcs of two different diameters.
T
No. 10 .-
cp
o. 11
- -~
o. 7
1nm
Agure 4.24 Atomically fesolved scanning tunneling microscope images of individual

single-walled carbon nanotubes. Reprinted with permission. from P. Moriarty, "Nanostructured
materials:' Rep. ?rag. Ph)'s.• 64. 369 (200!). ~ lOP Publishing Limited.
In Figs. 4.23 and 4.24, atomically resolved images of individual single-walled carbon
nanotubcs arc shown. The images were obtained by using scanning tunneling microscopy.
On the surface of the cylinders (tube images), it is secn clearly that the Ian ice structure
has a spacing between carbon atoms equal to 00 ~ 0.14 om. Such a (anice image makes
it possible to identify the chirality of the tube. Dashed arrows in Fig. ·t24 represent the
tube axis T and solid arrows indicate the direction of nearest-neighbor hexagon rows
H; compare this with Fig. 4.22(a). Tubes denoted by 10, 11, and I arc chirnl, whereas
lubes 7 and 8 have a zigzag and an armchair structure, respectively. Tube 10 has a chiral
angle 1ft = 7° and a diameter of 1.3 11m, which corresponds to the (11, 7) type shown in
Fig. 4.22(a). A hexagonal lattice is plotted on lOp of image 8 to prove that the non-chiral
armchair structure applies for this casco
In carbon nanotubes, an electron can propagate freely only in the direction along
the tube axis. Thus, electron or hole transport in these structures has the properties of
one-dimensiollal transport and the carriers can be characterized by a one-dimensional
wavevector k. As for any periodic structure, we can think of the electron energies as a
system of energy bands dependent on the waveveetor k. The energy bands of a single-
walled nanotube can be constructed from the energy bands of the graphene presented
in Fig. 4.22(a). For the case of a multi-walled carbon nanotube, the energy bands may
be constructed from the energy bands of a layer of graphite having a thickness equal
to the thickness of the wall of the nanotubc. Indeed, suppose that we find the energy
bands of the graphenc to be EN(K) with K = {K x . K},} being the two-dimensional
wavevector. Instcad of K., and K)" we can usc projections of K onto the tube axis, T,
and onto the wrapping vector, C, k and K c , respectively; i.e., the energy is EN(Kc. k).
However, along the C djrection there is a periodicity. Thus, we can use the so-called
cyclic boundary conditions, as we did in Section 4.4 to obtain the Brillouin zones. This
results in "quantization" of the K projection onto the vector C:
2rr N l
Kc= ~.35)
aov'3
with N 1 being an integer. For the electron energy. E(k), we obtain
(4.36)
That is, each of the initial bands of the graphene N is split into a number of one-
dimensional subbands N 1• In addition, the cha.racter of the splitting depends on the
wrapping type, i.e., on the values of (11. III) for the nanotube. Then, classification as
a metal, semiconductor, or dielectric can be made for these one-dimensional crystals.
For different (II. 111) values, distinct properties of the nanotube can be expected. A more
detailed theory predicts that when (11 - m) is divisible by three the single-walled carbon
nanotubes are of metallic nature; otherwise, they are semiconductors with bandgaps that
depend on the nanotube diameter:
(4.37)
where Ec- c is the binding energy of two carbon atoms, ao is the carbon-carbon distance,
and d is the diameter of the nanolube. Direct measurements support these considerations.
Specifically, for semiconducting tubes, an inverse linear dependence of the bandgap on
the nanotube diameter is confirmed, and the coefficient of proportionality is in good
agreement with theoretical expectations. It turns out that both types of carriers can exist
in those nanotubcs. i.e., both the electron and the hole can be responsible for electric
conductivity of the nlbes.
Similarly 10 the electronic properties, the mechanical propenies of nanolubes are

related closely to those of a single graphite sheet. Since the graphite sheet is very stitT
in the in-plane direction, one can expect nanotubes to have a similar stiffness constant
along the tube axis.
Indeed, it was found that, in contrast to the electronic properties, the elastic moduli of
nanotubcs are almost insensitive to the tube diameter and the chirnl angle. Their values
are comparable to those of diamond or ofa graphite sheet. In general, nanotubcs are very
stiff. For comparison, their elastic moduli are as much as five limes larger !.han those of
steel!
The response of the nanotubes to a large deformation is also remarkable. Indeed, most
hard materials fail with a strain ~ (see Eq. (4.26)) of about I% or even less, because of
generation of defects. The carbon nanohlbes can sustain up to 15% tensile strain before
fracture. Together with high stiffness, this provides a critical tensile strength of a single
nanotube as much as 300 10 400 times larger than that for steel.
In conclusion. carbon nanotubes represent a new class of nanostructures, which dif·
fer from traditional solid-state device structures. The nanotubes can be fabricated with
good control of their basic properties. including the electron bandstructure. They can
be fabricated as semiconductors either with electron conductivity. or with hole con-
ductivity. They can be contacted to metals and various types of processing techniques
have been shown to be viable for the fabrication of electronic devices from carbon
nanotubes.
Following the discovery of carbon nanotllbes, it was recognized that this kind of
perfectly organized l1anostrllcture should not be limited to carbon. It was found that for-
mation ofnanotllbes is a generic property of materials with anisotropic two-dimensional
layered structures. The structures of this type are called inorganic IWflOfllbe structures;
examples include WS 2• MOS2, V20 S , and B nanotubes. The study of these novel struc-
tures has led to the observation of a few interesting properties and promises potential
applications, particularly in nanoelectronics.
Buckyball fullerenes
Now we consider the so-called bUcAJlbaflJulferenes which represent lhe fourth major
form ofpure carbon. The previously discussed diamond and graphite and the well-studied
carbon nanotubes arc the other three.
The buckyball fullerene is a molecule with the chemical formula Coo and is one of
the best-known nanoscale objects in nanoscience. Figure 4.25 depicts a buckyball. It
consists of 60 carbon atoms occupying equivalent sites. Each atom is bonded to three
other atoms. Single and double C-C bonds occur. The lengths of the two types of bond
are 0.146 nm and 0.140 nm, respectively. That is, these bonds are practically identical.
In Fig. 4.25 they are represented by light and dark-shaded lines. To form almost an ideal
buckyball, nearest neighbors of carbon atoms are organized in pentagons and hexagons.
Every pentagon in the case struchlre is surrounded by five hexagons. The truncated
icosahedron structure shown in Fig. 4.25 has 12 pentagonal faces and 20 hexagonal
faces at 60 vertices with a C atom at each vertex.
pentagon
hexagon
Figure 4.25 A schematic drawing of the C60 buckyball (fullerene). Single and double bonds are
shown by light and dark shaded lines.
Because of the near-spherical shape of the truJlcated icosahedron, we can describe the
electronic energy states of the buckyball fullerene by using the classification of quantum
states developed for a spherically symmetric potential (Section 3.3) and the simplest
atoms (Section 3.4). As was found there. the electron cncrgy states can be classified by
angular momcntum quantum numbers (orbital mUllbcrs I). Ln the C60 fullercnc, the
total number of valence electrons is 240. Of them, 180 electrons are involved in
the relatively deep· lying a-bonding energy level, so that the electronic properties of the
fullerene are determined primarily by the remaining 60 electrons involved in 7(·bonds.
A total of 50 electrons may be accommodated in quantum states with orbital num-
bers from I = 0 to 1= 4. With two-fold spin degeneracy, this gives the following level
populations: 2 + 6 + 10 + 14 + 18, as illustrated by Fig. 4.26. TIle remaining 10
electrons fill the energy level with the orbital number I = 5. Actually, lhe total number
of these states is 2 x (2 x I + I) = 22. Now, we rccall that strictly speaking a trun-
cated icosahedron differs from a sphere and has a lower symmetry. On analyzing this
lower synmletry, one finds splitting of I-stales into three groups. as shown in Fig. 4.26
(h tu . tl u. and tl g denote different types of symmetry). The lowest five-fold (excluding
spin degeneracy) levels accommodate the remaining 10 electrons and can be designated
as the highest occupied molecular orbitals. Two other groups of 1=5 states are lifted
considerably; together with the I = 6 states they create the lowest unoccupied molecular
orbitals.
Conceptually, the electronic properties of an individual fullerene molecule can be
modified via replacement of a C atom with an atom having a greater or smaller number
of valence electrons. Such a process can be thought of as a kind of doping. An example
is azofuHcrene with the chemical formula CS9N. where one C atom is replaced by a N
atom.
I
5
6
6 Ilg (LUMO + 1)
5 Ilu (LUMO)
5 ***** hlu (HOMO)

4*********
3 ****
3 ***
2 *****
1 ***
o *
Figure 4.26 A diagram of lhe energy states of 60 valence electrons in the fullerene molecule.
After P. Moriarty, "Nanoslruclured materials:' Rep. Pl'og. Pltys.• 64, 355 (2001). © lOP
Publishing Limited.
The existence of unoccupied molecular orbitals for C60 explains effects ofinlcrmolec-
ular interaction ofthe buckyballs. Forexamplc, twoC60 molecules, being neutral systems,
attract each other through polarization of electron clouds. Such an attraction is caused by
van def Waals forces. Owing to these forces. two fullercncs compose the dimer C6Q-C60
with the coupling estimated to be about 0.27 cv. Similar van der Waals forces pro-
vide for interaction and adsorption of the fullerenes onto surfaces of various materials.
Figure 4.27 depicts the fullerene molecules adsorbed onto a Si surface. Since the van der
Waals forces are relatively weak, the fullerene molecules readily diffuse on the surface
and, in large concentrations, they can form hexagonally close-packed islands, as shown
in Fig. 4.28.
Then, the buckyballs themselves form face-centered cubic crystals (at room tem-
perature) with large cohesive energy in this crystal (.. . . . 1.6 eV per molecule). Inter-
estingly, the lattice constant of the C60 crystal has the unusually large value of
1.42 nm.
Summarizing, the fullerenes are natural and very stable nano·objccts manifesting a
number of interesting physical and chemical properties, which can be controlled and
used in nanoscience applications.
4.9 Closing remarks
We began this chapter with the dcfinjlion of crystalline materials. We introduced two
very important components of crystals, i.e., the electron subsystem and the crystalline
lattice. The electronic applications ofa material arc determined primarily by its electronic
4.9 Closing remarks 105
-10nm
FIgure 4.27 An image offullerene molecules adsorbed on a silicon surfacc (al low
concentration). Reprinted with pennission. from P. Moriarty." anostructured materials:·
Rep. Prog. Phys., 64. 306 (200 I). lOP Publishing Limited.
Agure 4.28 At large concentrations, the fullerenes foml a hexagonal structure on a surface.
Reprinted with permission. from P. Moriarty, ';Nanostructured materials;· Rep. Prog. Phys., 64,
358 (2001). () lOP Publishing limited.
properties. As a result of the periodicity of crystals. one can interpret the electron energy
in terms of the energy bands. Depending on thc relative positions of the energy bands
and their filling by electrons, a material can be described as a dielectric, semiconductor.
or metal. We found that. despite the interaction of the electrons with the great number of
atoms and ions composing a crystal. one can anribute a quasimomentum (a wavevcctor)
to an electron and tbat. frequently, within the actual range ofthe wavcvectors, the electron
energy can be approximated by a simple dependence on W'<lvevector. Several new notions

that are necessary to describe electrons in these cases were introduced: (I) the positions
of the energy minima, (2) the bandgap, and (3) the effective masses.
We analyzed scmiconductor alloys and learned that for many cases of practical intcrcst
the alloy energy spectra can be described in a manner similar to that used to describe
purc crystals. These alloys may be "cngineered" to produce considerable variations in
the electron parameters.
Next, we explained the formation of discontinuities in both valence and conduction
bands when two materials are brought together to form a heterojunction. This effect
provides the practical basis for modifying electron energy spectra, and it is particularly
useful for spatial modulation of the potential profiles experienced by electrons and holes
as well as for the creation of various artificial heterostructure~based nanostructures with
energy barriers; these nanostructures include quantum wells. quantum wires, quantum
boxes, and supcrlanices.
By analyzing the state of the art in the fabrication of hcterostructurcs, it was estab-
lished that high-quality heterostructurcs can be produced using materials with similar
crystal properties, for example, materials from the same group. The ratio of lan-ice con-
stants of the two materials is a critical parameter for such heterostructures. If these lattice
constants almost coincide, as in the case oflaltice~matched materials, one can produce a
heterostructure without strain due to the absence oflallice mismatch or misfit imperfec~
tions. An example ofsuch a lattice-matched system is the AIGaAs/GaAs heterostructure.
For lattice~mismatched materials we have found that only thin layers can accommodate
the lattice mismatch and retain near~perfect crystalline structure. The resultant structures
arc pseudomorphic straincd layers. The strain in pscudomorphic hctcrostructures leads
to a sct of ncw phenomena. In particuJar, it affects thc energy spcctra in the strained
layers. An example is the Si/Ge heterostructure.
Finally, we considered the structural and electronic properties of novel nanosized
objects such as carbon nanotubes and carbon buckyballs.
More information on crystal symmetry, electron energy spectra, and the general theory
of strain effects can be found in the following books:
C. Kittel, Quantum TheOlY 0/ Solids (New York, John Wiley & Sons, Inc., 1963).
1. Ipatova and V Mitin, Introduction (0 Solid-State Electronics (New York, Addison-
Wesley Publishing Company, 1996).
A briefhistory ofsemjconductor hetcrostructures and clear motivation fortheir imple-
mentation is given by Professor Herbert Kroemer in his Nobel Lecture:
H. Kraemer, "Quasielectric fields and band offsets: tcaching electrons new tricks;'
Rev. Mod. Pllys., 73. 783 (2001).
Detailed treatments ofsemiconductor alloys and hetcrojunctions as well as reviews of

methods of their fabrication and doping are presented in the following references:
J. Singh, Physics o/Semiconductors and Their Heterostructures (New York, McGraw-

Hill, 1993).
4.10 Problems 107
v. V. Mitin, V A. Kochelap, and M. A. Stroscio, Quantum Heterostructures (New

York, Cambridge University Press, 1999).
The following publications arc devoted to calculations and discussions of electron

energy spectra and wavcfullctions in various quantum semiconductor stnlctures:
T. Ando, A. B. Fowler, and F. Stern, "Electronic properties of two-dimensional sys-

tems,'" Rev. Mod. Ph)!s., 54, 437 (1982).
G. Bastard, WOlle Mechanics Applied to Semiconductor !-!eterostrUCllIres (New York,
Halsted Press, 1988).
C. Wcisbuch and B. Vinter, Quantum Semiconductor Srrucfures (San Diego, CA,
Academic Press, 1991).
A detailed description of carbon nanotubcs and carbon buckyballs is presented in the

following recent publications:
M. S. Dresselhaus, G. Dressclhaus, and P. C. Eklund. Science of Flillerenes alld

Carbon Nano!lIbes (San Diego, CA, Academic Press, 1996).
P. Moriarty, "Nanostructural materials," Rep. Prog. Phys., 64, 297 (2001).
4.10 Problems
1. Using the lattice constant of silicon, a = 5.43 x 10- 8 cm, and the fact that the
number of Si atoms per unit volume, a], is eight, calculate the number of atoms per
I cm] and the density of the crystalline silicon (silicon '5 atomic weight is 28.1 g mole-I).
2. Estimate the volume of the first Brillouin zone in k·spacc for a simple cubic lattice
with the lattice constant a = 5 x 10- 11 cm. Assume that the average energy of the c1ec·
trons is 3k B TJ2, where kB and Tare Boltzmann's constant (k a = 1.38x 10- 2] J K- 1)
and the ambient temperature, respectively. Estimate the volume occupied by electrons
in k-space at T = 300 K with the effective mass 11/ = 0.11110, with 11/0 being the free-
electron mass. Compare these two volumes and discuss whether Eq. (4.17) is valid for
electrons with the parameters specified previously.
3. lllree electrons of the same energy are placed into three different energy valleys
of silicon. The valleys are located at fj. points of the [100]-, [010]-, and [OOI]-axes.
Asswning that all three electrons move along the same direction, say [100], and lIsing
Eqs. (4.17) and (4.19), find the ratio of the velocities for these electrons.
4. Consider a valence band consisting of light- and heavy-hole branches. Assume that a
heavy hole with energy E is transferred to a light-hole state with the same energy E. Find
the ratios ofille quasimomenta and the velocities of the hole in the initial and final states.
5. For AlxGal __~As alloy, find the composition having an energy bandgap equal to
2 eV using Table 4.6. For this alloy, determine the effective masses in the r and X
valleys using Table 4.7.
6. Assume that for some applications it is necessary to use a film of InGaAs of high
quality, which can be grown on an InP substrate. By using Eq. (4.20) and data presented
in Table 4.8, find the lattice·matched composition of this alloy, its lattice constant, the
energy bandgap, and the wavelength of the light corresponding to this bandgap.
7. Assume that the conduction·band offset for an AlxGal_xAs/GaAs heterojunction is

60% of the difference of the bandgaps of these materials. Find the composition of the
AlGaAs layer necessary for the resulting heterojullction to have an energy barrier for the
electrons equal to 0.3 eV. Calculate the energy barrier for the holes.
8. On the basis of the values oflattice constants given in Table4.8, explain why it is fea-
sible to grow stable AI.~Gal_xAs/GaAs and InxAll_xAs/hlyGal_yAs heterostructures;
explain why it is difficult to grow stable hctcrostructures of GaP/SiC and hlP/SiC.
9. Find the numbers of hexagons per unit cell in zigzag and armchair nanotubes. Cal-
culate the radii of (7, 8) and (11, 10) nanotubes.
5 Growth, fabrication, and
measurement techniques for
nanostructures
5.1 Introduction
Having reviewed the basic properties of materials exploited in nanoeleclmnics, we shall

now study the principal methods ofhigh-qllality material growth and nanodevice fabrica-
tion. Methods forthe growth ofperfect materials with controllable properties are critically
important for nanostrllchlre fabrication. Indeed, stringent requirements must be met for
the growth ofcrystals for nanosize devices. These requirements include Illany factors and,
first of all. ultra-high quality and purity. both controlled within extremely close limits.
The following examplcs illustratc the term "ultra-high quality." For Si crystals used in
nanodevices, conccntrations of controlled impurities currently reach concentrations of
less than one part in tcn billion (I in 1010). For the case of Ge, this number is in the
range of I in 10 13 _10 14 . The quality of a silicon crystal being used for nanoelectronics
can be characterized in tcrms of the density of defects: thcy must be limited to several
tcns per I m2 (!) of the Si wafer according to the Semjconductor Road Map, that was
discussed in Chapter I. The basic methods of growth of perfect crystalline materials and
Illultilayered heterostructures we will discuss in Section 5.2.
To fabricate a nanostructure and a nanodevice two approaches can be undertaken.
The first is based on a previously grown perfect material with further processing. This
includes a number of fabrication stages and methods (nanolithography, etching, implan-
tation, selectivc doping, etc.). In Scction 5.3 wc rcview thcse mcthods. The second
approach exploits spccial rcgimes of material growth. when nanostructures are formed
spontaneously due to the growth kinetics. Such a growth regime can control size.
shape, and other properties of the nanostruchlres. Details of this approach are given in
Section 5.4.
Progress in the refinement of fabrication techniques for making nanostructures
depends on the great improvements made in characterization methods. In particular,
composition and dopant distribution, lattice strain, and other parameters within nanos-
tructures must be known with atomic-scale precision. Currently, the manipulation of a
single atom (ion) in a solid is possible. In Section 5.5, we shall present the most impor-
tant characterization techniques, such as atomic-force microscopy, scanning tunneling
microscopy, transmission electron microscopy, and others.
The previously mentioned approaches for the production ofnanostructures and nano-
electronic devices actually represent "evolutionary" improvements in the growth and
110 Growth, fabrication, and measurement techniques
processing methods applied previously in microelectronics. anoseale objects like

carbon nanotubcs and biomolecules require, in general, other techniques for produc-
tion. These innovative techniques are highlighted in Sections 5.6 and 5.7.
Advances in nanotechnology require the utilization of methods and concepts from
almost every area of science and engineering. Synthetic chemistry and even biology
have much to offer for this emerging field. Some fundamental concepts coming from
these fields can successfully be exploited. These include chemical and biological meth-
ods of surface nanopatterning, and preparing nanostructured materials with predefined,
synthetically programmable properties from common inorganic building blocks with
the help of DNA interconnect molecules. The basic ideas related to these chemical and
biological approaches are discussed in Section 5.8.
The great technological advances brought about in mainstream microelectronics and
nanoelectronics can be used for the fabrication of another class of nanodevices thai
employs both electrical and mcchanical propertics of nanostructures. This new genera·
tion of dcvices is commonly called lIanoelectl'Omechallical systems (NEMSs). Indeed,
a strong enhancement of coupling between electronic and mechanical degrees of free-
dom appears on the nanometer length scale. This results in a new class of devices
that includes nanomachines, novel sensors, and a variety of new devices functioning
on the nanoscale. In Section 5.9 we study the methods of fabrication of this class of
nanodevices.
5.2 Bulk crystal and heterostructure growth
Though technological methods and especially regimes of growing various types of crys-
tals are generally different, they have a lot in common. Here we consider the common
steps in growing pure materials using Si technology as an example.
Single-crystal growth
The following three steps arc necessary to produce high-quality silicon crystals:
(i) production of metallurgical-grade silicon (impurity level ~5 x 10 16 cm- 3 ); (ii)
improvement of the laner matcrial up to electronic-grade silicon (the level of impu-
rities is reduced to ~5 x IOU cm-3 or less); and (iii) conversion 10 single-crystal Si
ingots.
Metallurgical-grade silicon is lypically produced via reaction of silicon dioxide (Si0 2)
with C in the form of coke:
SiD, + 2C .... Si + 2CO. (5.1)
which requires very high temperature (::::::1800 °C). Coke is a coal from which most of the
gases have been removed. The silicon obtained at this step is nol single-crystalline and
is not pure enough for electronic applications, though il is good for some metallurgical
applications such as the production of stainless sleel.
5.2 Bulk crystal and heterostructure growth 111
Pull
t
>Wafers
_Growing
cry tal
Molten Si
Figure 5,1 The Czochralski method for the growth of bulk semiconductors.
Further reductions in impurities can be achieved by carrying oul the following reaction
of the silicon with dry Hel:
Si + 3HCl -+ HSiCl, + H2 • (5.2)
Now we obtain the trichlorosilanc HSieh. which is typically in a liquid state with
a boiling point of 32°C. Simultaneollsly with I-ISiCI J • other chlorides of impurities,
such as FeCht are formed. Since their boiling points are different, the simple fractional
distillation technique can be applied: the mixture ofHSiCb and other impurity chlorides
is heated and thcn condensed in a series ofdistillation towers at appropriate temperatures.
By this technique HSiCI J is separated from impurities. The following reaction with H2
then converts the trichlorosilane into highly pure electronic-grade silicon:
2HSiCI J + 2H, -+ 2Si + 6HCI. (5.3)
The pure Si obtained by this process is still polycrystalline.

The final proccss, which converts polyerystalline silicon into single-crystal Si ingots,
is based on the Czochralski mcthod. In this method, a seed Si crystal provides a template
for growth. First, tllis seed crystal is lowered into the molten Si material. (The melting
point of Si is 1412 0C.) Then it is raised very slowly so that the molten material touching
the seed crystallizes as the seed is withdrawn from the molten material. Rotation of the
seed crystal, stabilization ofthe temperature field, and other tricks are used to grow highly
homogeneous ingots. The Czoehralski method is illustrated in Fig. 5.1. Importantly, this
technology facilitates doping in the course ofcrystal growth. Indeed, one can intentionally
add precise quantities of impurities (dopants) into semiconductor melts to provide for
regions of crystallization having the desired doping concentrations. This technique is
used widely in growing silicon, germanium, and, with some modifications, compound
semiconductors.
As the single-crystal ingot is grown, it is mechanically processed to obtain wafers of
thicknesses of hundreds of micrometers, as shown schematically in Fig. 5.1. The wafers
are used subsequently for producing individual devices, or integrated circuits, or for the
GaAs
--
Substrate
~ Molecular beams
1-
I
\/ \\ Shutters
I
I
\ I
I >::
....I....
Ga
Figure 5.2 Thc MBE mcthod for thc growth ofGaAslAIGaAs hctcrostnlcturcs.
fabrication of much more complicated structures. In the following discussion, we focus

on the technique of growing single-crystal layers above a substrate wafer, which is called
epitaxial growth, or just epitaxy.
Epitaxial growth
Fabrication of a crystal layer upon a wafer of a compatible crystal makes it possible 10
obtain very well~controlled growth regimes and to produce high~quality crystals with the
desired crystalline orientation at temperatures typically well below Ihe melting point of
the substrate. During the epitaxial growth, several methods of delivering the necessary
atoms to the growing layer can be lIsed. The most developed methods are molecullll'-beam
epitaxy (MBE), chemiclll-vapordepositioll (CVD), and Iiqliid-phllseepitaxy(LPE). Here
we shall discuss M BE and CVD in more detail.
Molecular-beam epitaxy
The MBE method can be realized in a high vacuum, where molecular or atomic beams
deliver onto a substrate the necessary components for growing the desired crystalline
layer. For example. suppose that we want to grow all AIGaAs layer on GaAs. Then,
the substrate will be GaAs and the atomic beams are fluxes of the elements AI, Ga,
and As, as well as beams of dopants (typically, Si is used for n-doping and Be for
p-doping). Sources of the elements arc contained in separately heated chambers. The
evaporated clements form beams, which are separately and closely controlled, collimated,
and directed onto the substrate surface, as illustrated by Figs. 5.2 and 5.3. Typical
flux densities in the beams are of 10 14_10 16 atoms cm- 2 S-I. The substrate is held
at relatively low temperature (:::::::600 °C for GaAs), while dcnsities of the components
in the beams are large. This provides effective growth of the layer. A slow growth rate
(::::=: 1 monolayer per second), which is often referred 10 as layer·by·laycr growth, results in
5.2 Bulk crystal and heterostructure growth 113
Rgure 5,3 A typical MBE chamber. Used with permission from

www.ccc.cmu.cdu/......dwg/rescarchjmbcchambcr.htmi. © 2007. ECE.
the growth ora high-quality layer. By controlling shutters for each beam, one can produce
abrupt changes in crystal compositions and doping concentrations on the scale of one
monolayer.
Chemical-vapor deposition
This epitaxial method allows one to realize a low.lcmperature growth regime and to
use high-purity chemicals for delivering the necessary atoms for growth of a crystalline
layer. The layers can be grown onto a seed crystal or substrate from mixtures of chemical
vapors containing both semiconductor elements and dopants.
Gas bame
-------- - -
j} W Wafers W 0
Quartz chamber
0
0 0
0
•• •
0
0 0
. 0 0
Radio-frequency 0 0 SUSCCplOT
source 0
0
0 ••• 0
0
0
0
0
0
j)
•• • 0
0
0
0
Pedestal
Figure 5.4 The CYD method for the growth orhcterostructures.
The chemical-vapor deposition is conducted in a reaction chamber called a reactor,

as il1ustrated by Fig. 5.4. In the reactor, a typical pressure of chemicals is:::::: 10 4 Pa and
healing is achieved by power from a microwave radio-frequency source.
In the case of growth of Si layers, several different gases containing Si atoms can
be used. They include silicon tetrachloride (SiCI.,), silane (SiH 4 ). and dichlorosilanc
(SiH 2 C1 2 ).ln the lise ofsilicon tetrachloride, the following reaction with hydrogen occurs:
SiCI, + 2H, --> Si + 4HCI.

The reaction can be conducted at temperatures in the range of 1150-1250 °C. In the
case of using silane and dichlorosilane, the reaction can be conducted at even lower
temperatures (1000-1100 °C). These temperatures are well below the melting point of
Si. Thus, these reactions release atoms ofSi, and the relatively low-temperature regimes
provide efficient crystal growth onto the seed.
For growth of Am-B v compounds, the following reactions are used:
These reactions take place at temperatures of ~600-700 0c. Dopants can be delivered,
for example, by silane (n-doping) or diethylzine (p-doping).
It is important that epitaxial methods can be applied to produce new materials that are
difficult to grow by other methods. Examples are wide-bandgap nitrides of the group III
elements. These include InGaN and AIGaN compounds.
In conclusion, epitaxial growth on the basis of CVD and MBE methods has become
the dominant technique for the fabrication of perfect multilayered crystals of nanoscale
thicknesses. Extremely high uniformity oflayers has been demonstrated to be achieved by
these epitaxial methods. Both group IV elements and IU-V compounds are successfully
5.3 Nanolithography, etching, and other means 115
grown with thickness control of the order of one monolayer. Different types of doping -
uniform doping, modulation doping, and delta-doping - are realized with high accuracy.
Since in the chemical reactor the partial pressures of chemicals are much higher than the
pressure in the molecular beams of the MBE method, the rate of crystal growth realized
in the CVD mcthod is higher than that of MBE. The former may be used in industrial
production, while the latter is rather well suited for research laboratories.
5.3 Nanolithography, etching, and other means for fabrication

of nanostructures and nanodevices
In the previous section, we studied methods ofgrowing high-qual ity single-crystal wa fers
and crystalline multjlayered structures. We found that thicknesses of the layers may be on
the nanometer scale. However, 10 produce an individual device or electric circuit scaled
down to nanosize in two, or three, dimensions. one needs to exploit additional mcthods.
First of all, on a wafer it is necessary to produce patterns corresponding to feahlres of the
nanodevice or circuitry. This may be done by using one of the so-called nanolithography
methods.
Let us start with photolithography. This method involves the generation of a reticle,
which is essentially a transparent quartz plate with the necessary pattern. Opaque regions
on the reticle are made up of an iron-oxide layer, which absorbs ultraviolet (UV) light.
A pattern on the reticle is usually made by a computcr-controlled electron beam, which
moves as prescribed by pattern-generation software. An electron-beam-sensitive material
(the electron-beam resist) is placed onlO the iron-oxide-covercd quartz. The resist is
exposed selectively by the electron beam and then the exposed material (positive resist)
is removed. Finally, the iron·oxide layer can be selectively removed by etching to generate
the desired pattern on the quartz plate.
During the next step, a thin uniform layer of photoresist is deposited onto the wafer
surface. There are two types of photoresists: positive and negative. For positive resists,
the resist is exposed with UV light wherever the underlying material is to be rcmoved.
In these resists, exposure to the UV light changes the chemical structure of the resist
so that it becomes more soluble. The exposed resist is then washed away by developer
solution,leaving windows of the bare underlying material. The mask, therefore, contains
an exact copy of the pattern which is to remain on the wafer. Negative resists behave in
just the opposite manner. Exposure to the UV light causes Ihe negative resist to become
polymerized, and thus more difficult to dissolve. Therefore, the negative resist remains
on the surface wherever it is exposed, and the developer solution removes only the
unexposed portions. Masks used for negative photoresists, therefore, contain the inverse
(or photographic "negativc") of the pattern to be transferred. Figure 5.5 shows the steps
involved in photolithography, as wetl as the pattern differences generated by the use of
positive and negative resists. Positive resists are now the dominant type of resists used
in fabrication processes.
To transfer the patterns onto the wafer surface, the mask should be aligned with the
wafer. Once the mask has been aligned accurately with the pattern on the wafer surface,
Illuminating radiation
!!~!!~!!~!!
Quartz plate b~;===;~;:::=:;;;;~I- Mask
---
lon-oxide layer
Imaging system
__ Photoresist
I I ..... .....- Film to be pattemed
rIluminated areas
Substrate
Negative resist: / ~ Positive resist:
exposed rendered Develop exposed is
insoluble removed
z~31;j7 Etch pattern I

r. ..
Figure 5.5 Basic lithographic processes.
the photoresist is exposed through the pattern on the mask to high-intensity UV light.
One of the exposure methods, the contact method, is shown in Fig. 5.5. In this method,
the photoresist is exposed to UV light while the wafer is in the contact position with the
mask. As a result of the direct contact between resist and mask, very high resolution is
possible in contact printing.
One of the last steps in the photolithography process is development. The results
obtained after exposure and development arc shown in Fig. 5.5 both for negative and for
positive resists.
Actually, photolithography processes have been known for decades. To use them for
nanostructures. we have to understand the limitations imposed by the wave nature of
light. Indeed, as we discussed in Chapter 2, simple geometrical ray optics is applied for
dimensions greater than the wavelength of light. The wave effect - diffraction of light -
restricts the application of photolithography to minimum dimensional scales of abollt
the wavelength of light. Thus. shorter wavelengths are preferable. Examples of some of
the shorter wavelengths in lise arc 0.365 ~.m for UV mercury lamps and 0.193 ~t.ln for
ArF excimer lasers. With these UV sources, it is possible to achieve linewidths of about
0.25 Il-m and 0.15 ~m, respectively. Further penetration into the deep-UV region appears
to be extremely difficult.
The previously mentioned diffraction limit is much smaller for X-rays, electron beams,
and ion beams. Thus, advances in nanolithography are occurring as a result of the use
of these short-wavelength beams. For example, electrons with an energy of 10 keY have
•
a wavelength of about 0.1 A; i.e., less than the lattice constant's of any crystal. Now, the
5.3 Nanolithography, etching, and other means 117
Plasma radicals and ions

-- ,
Etch gas
) 1 1 t- o p ump
.......Anode (Chamber)
\
wirer
-
-
Catiilide
Rf
power supply
Figure 5.6 Reactive-ion etching.
ultimate linewidth is determined by interaction with the photoresist layer. In principle,

it is possible to reach linewidths below 0.1 ~m for direct writing with an electron beam
onto the photoresist.
Etching
As soon as a resist pattern is formed, an etching method is generally used to pro·
dllcc the desired stnlcturcs as shown in Fig. 5.5. There are many etching methods.
A widely used method is wet chemical etching. For example. dilute HF is used to
etch a SiDl layer covering silicon. The HF reacts with SiDl and does nol affect the
photoresist or silicon. That is, this wet chemical etch is highly selective. However,
the ratc of etching is the same for any direction, lateral or vertical, so the ctching is
isotropic. Using an isotropic etching technique is acceptable only for relatively large
structures. For nanosize structures, anisotropic etching with faster vertical etching is
preferable.
Anisotropic etching generally exploits a physical process, or some combination of
both physical and chemical methods. The besl~known method of anisotropic etching is
reactive-ioll etching. Reactive-ion ctching is based on the use of plasma reactions. This
method works as follows. An appropriate etching gas, for example a chlorofluorocarbon,
fills the chamber with the wafers. The pressure is typically reduced, so that a radio-
frequency (RF) voltage can produce a plasma. The wafer we want to etch is a cathode
of this RF discharge, while the walls of the chamber are grounded and act as an anode.
Figure 5.6 illustrates a principal scheme for the ion-etching method. The electric voltage
heats the light electrons and they ionize gaseous molecules, creating positive ions and
molecular fragmel1ls (so-called chemical radicals). Being accelerated in the electric field,
the ions bombard the wafer normal to the surface. This normal incidence of bombarding
ions contributes to the etching and makes the etching highly anisotropic. This process,
unfortunately, is not selective. However, the chemical radicals present in the chamber
give rise 10 chemical etching, which, as we discussed, is selective. Now we see that the
method combines both isotropic and anisotropic components and can give good results
for etching on the nanoscale.
Doping
As we discussed in Section 4.2, a perfect semiconductor is a djclectric at low tempera-
tures: the valence band is completely filled with valence electrons, while the conduction
band is totally empty. When the crystal temperature rises, some electrons can be excited
into the conduction band, which results in the creation of a pair: an electron in the con-
duction band and a hole in the valence band. Thus, for the electron concentration,". and
the hole concentration, P, we obtain
" = p = "i(T)·
Here, II , (T) is the temperature-dependent intrinsic density of electrons and holes. Such
a perfect semiconductor is itself called an intrillsic material. In practice. the intrinsic
concentrations are small and can not be controlled effectively.
Thus, it is important to find a method to create the electron orland hole concentrations
needed for each particular device application. The Illost common method of varying
electron and/or hole concentrations in a crystal is the process of adding impuritics to the
material; i.e., the doping pmcess. The process is based on the following physical picture.
When an impurity is added to a semiconductor, additional energy levels are contributed by
the impurity to the energy structure of the crystal. Many of these additional energy levels
fall near the bandgap energies. Here, we consider two especially important cases: (I) the
impurity levels are close to the conduction band and (2) they are close to the valence b'md.
In the former case the impurities can be dOllors ofelectrons to the conduction band, while
in the latter case they can be acceptors; i.e., they will accept electrons from the valence
band and thereby generate holes in the valence band. For example, impurities from group
V of the periodic table of elements (P, As, Sb) arc donors and impurities from group 111
(B, AI, Ga, In) are acceptors for group IV semiconductors. The energy "distances" of
donor levels from the conduction band and of the acceptor level from the valence band
are much smaller than the energy bandgap. This promotes thermally induced ionization
of these impurities even at low temperatures and the creation of conducting electrons
or holes. Ln Fig. 5.7, the population of energy states in semiconductors with donors and
acceptors is illustrated for zero and finite temperatures.
We can perform the following simple estimates of energy levels of donors and accep-
tors. Consider first, for example, an atom from group V embedded in a semiconductor
material of group IV Such an atom has five valence electrons, of which four can par-
ticipate in formation of covalent bonds with neighboring group IV atoms. But the fifth
electron is an extra electron. We can consider this electron to be moving around a positive
ion, i.e., Ihe impurity can be thought of as a "hydrogen atom" embedded in a dielectric
medium. Ifthe radius of the electron state for this atom is largc. we can suppose that the
electron has the effective mass "'., which is characteristic for the conduction band. By
5.3 Nanolithography. etching, and other means 119
(a)
E, I I I I I
Ed Ed
, £,
•••••••••••••••••• ••••••••••••••••••
•••••••••••••••••• • •••••••••••••••••
•••••••••••••••••• • •••••••••••••••••
I I
" " " I I I I •••• ".
T=O K T= 50 K
(b)
E,t--------j £,
E,
E,
•••••••••••••••••• ••••••••••••••••••
••••••••••••••••••
••••••••••••••••••
•• •••••••••••••••••
•••••••••••••••••
T=OK T= 50 K
Agure 5.7 Doping of semiconductors: (3) donor doping lind (b) acceptor doping.
applying the result's of Section 3.4 that were obtained for the hydrogen atom, we find the
ground-state energy of such a "hydrogen".like state of the impurity:
(5.4)
with E'o and € being the permittivity of free space and the dielectric constant of the mate-
riaJ, respectively. For an acceptor atom, we may apply the same approach, considering a
hole moving around a negatively charged ion. Equation (5.4) will be valid if the effective
mass of the electron in the conduction band, m·, is replaced by the effective mass of a
hole, mh, in the valence band: m· ~ my. Typical energies of donor and acceptor levels
in silicon, for example, are about 30-60 meV from the conduction and valence bands.
and the ionization occurs at temperatures above several tens of degrees Kelvin.
Similarly to the previously considered case of materials of group IV, in III-V com·
pounds the atoms from group VI (S, Se. Te) arc donors when they occupy sites of group
V atoms. Atoms from group II (Be, Zn, Cd) act as acceptors when substiruting group
III atoms. A more complicated case occurs when doping is accomplished with atoms
of group IV; for example, with Si or Ge. These impurities are called amphoteric, which
means that they can act either as donors or as acceptors, depending on the sites they
occupy. If such an atom occupies a group III site, it brings an additional electron and
acts as a donor. When occupying a group V site, the atom accepts an electron and acts
as an acceptor. In GaAs, typically, Si impurities occupy Ga sites; thus they are typically
donors. However, if during the growth process there are As vacancies. Si atoms can fill
these vacancies and serve as acceptors.
In conclusion, the doping of semiconductor materials provides a powerful tool
for realizing the desired type of conductivity - electron conductivity or hole
120 Growth. fabrication, and measurement techniques
conductivity - and fabricating structures with the desired values of electric resistance.
Doping methods provide a means for the control of the electronic properties over a
wide range of values. lllCSC methods are well suited for application in the fabrication of
semiconductor nanostructures.
5.4 Techniques for characterization of nanostructures
Progress in the fabrication, study, and use of nanostruelures would not be possible with-
out adequate techniques for the characterization of these structures. These techniques
shouJd allow one to determine the shape and geometrical parameters ofnanostructures,
the distribution of chemical composition, the strain fields, etc. Knowing all of these,
one can predict the elcctronic and optical properties which wiJlultimately be relevant in
applications. The questions of what the geometrical parameters of nanostmctures are,
and what their shapes are, can be addressed using atomic-force microscopy (AFM) or
scanning tunneling microscopy (STM). Cross-sectional variants of STM and transmis-
sion electron microscopy (TEM) are sufficiently powerful tools for the investigation of
buried structures. X-ray diffraction and some photolumincscence mcthods may also be
used to detcrmine size and shape, but in general AFM, STM, and TEM have better res-
olution. On thc other hand. X-ray diffraction is a very powerful tool for measuring strain
fields, defects, and imperfections. For the determination of statistically relevant proper-
ties, averaging over many individual nanostructures is necessary. STM and TEM often
do not facilitate the study of large enough areas of a sample to yield quantitative data.
Photoluminescence and diffraction techniques make possible the snldy of intrinsically
large ensembles ofnanostructures, so tbat a good statistical average is obtained automat-
ically, whereas the quantification of fluctuations is sometimes difficult. The X-ray and
optical methods are routinely applied for material cbanlcterization, and discussion about
them can be found elsewhere.
In tbis section, we shall present a brief overview of new techniques based on AFM,
STM, and TEM, which can be applied for a detailed characterization of nanostructures.
Scanning tunneling microscopy

This novel technique yields surface topographies and work-function profiles on an atomic
scale directly in real space. In Section 4.5, we explained thallhe removal of an electron
from the conduction band of a solid requires a certain amount of energy called the
electron affinity. For a metal or a doped semiconductor, when the conduction band is
partially filled. the energy needed to remove an electron is lower and it is called the
workjimctioll. Let us consider two conducting solids separated by a space. In terms of
classical physics, a transfer process of an electron from olle solid into another can be
thought of as an electron transfer over a vacuum barrier. The process requires additional
energy and because of this it has a small probability. In Section 3.3 we found that,
according to quantum mechanics, a particle can penetrate a classically forbidden spatial
5.4 Techniques for characterization 121
p
v.
B
:->d'-:;::-l_• --
p
r.A •• ·••·
••
••••• •
••
CllIltrO]
Unit
Figure 5.8 The principle of operation of the scanning tunneling microscope.
region under a potential barrier. This phenomenon was called funneling. Thus, electron
transfer between two solids can occur as a tunneling process through (under) the vacuum
barrier. Tunneling experiments have been performed for example, by llsing two metal
films separated by vacuum or a sol id-statc insulator (3 sandwich struchlre). Each of the
metal films can be considered as an electrode and when a voltage bias is applied to
these electrodes a so-called 111111leUng electric curren! is produced. This current can give
information on electronic properties, hut obviously lhe infonnation will be averaged
over the area of the metal-film surface. By appropriate shaping of one of the electrodes
spatial resolution of far smaller scales than that of sandwich structures can be achieved.
Since vacuum is conceptually a simple tunnel barrier, such experiments pertain directly
to the properties of the electrodes and their bare surfaces. Clearly, vacuum tunneling
offers fascinating and challenging possibilities to study surface physics and many other
related topics.
The principle of STM, which is based on electron tunneling, is st.raightforward. It
consists essentially of a scanning metal tip (one electrode of the tunnel junction) over
the surface to be investigated (the second electrode), as depicted in Fig. 5.8. The metal
tip is fixed to a rectangular piezodrive marked as p.r. Py and P=. The piezodrive is made
of a piezoceramic material. The tunnel current, h, is a sensitive function of the gap
between the tip and the surface, s; i.e., h <X VT exp(- Aq, 1/2 S ), where q, is the average
barrier height; the numerical value of A is equal to unity if 0, is equal to the work function for a metal
or doped semiconductor. With a typical value of <p of several e'Y, h changes by an
•
order of magnitude for every I A change of the gap, s, The control unit, CU, applies
a DC voltage, V=, to the piezodrive, P=, such that IT remains constant when scanning
the tip with PI and p)' over the surface. At constant function <p, V=(x.y) yields the
topography of the surface, that is .!(x. y), directly, as illustrated at a surface step in
Fig. 5.8, The curvature of the tip causes the abrupt step to appear to be smeared over
a distance o. For a constant tunneling current, changes in the work function, <p, are
compensated by corresponding changes in the distance, s. Thus, a lower work function

at a contamination spot C would be observed as an additional surface structure denoted
by B in the Fig. 5.8. These "work·function·induced stnlctures" and true structures can,
however, be separated by the following method. Let the gap, s, be modulated by fj,s while
scanning at a frequency higher than the cut-off frequency of the control unit. Thus, the
modulation is no longer compensated by the feedback loop ofthe control unit. As a result,
the current will be modulated by 6. h. Then, the ratio id = fj,(ln h)/ 6.s ~ 4>1/2 directly
gives the work function revealing the spot C. in the simple situation shown in Fig. 5.8.
Since V,;, h, and id can be measured, the topography and the work function can be
reconstructed.
This principal scheme of tunneling microscopy provides (i) stability ofa vacuum gap
• •
in the sub-A range and (ii) a lateral resolution in the A range. This requires excellent
vibration damping and very sharp tunnel tips. The first requirement is met by using
highly developed and clever mechanical means. In fact, it is possible to lise bungee
cords if they are properly placed and have the desired elastic properties! Tunnel tips used
nowadays are typically made of tungsten or molybdenum wires with tips of overall radii
< I ~m. However, the rough macroscopic grinding process creates many rather sharp
minitips. The tunnel current is extremely sensitive to the vacuum gap, s: this is why
the millitip closest fO fhe sample defines the whole current through the tip. Actually,
the lateral resolution is given by the width of the tlillnel channel, which is extremely
narrow. Additionally, focussing of the tunneling current (in addition to the geometrical
one) occurs due to a local lowering in height of the tunnel barrier at the apex of the lip .
•
At present, the resolution of scanning tunneling microscopy reaches 0.05 A vertically
•
and well below 2 A laterally.
Scanning tunneling microscopy is subject to some restrictions in application: only
conductive samples can be investigated, and measurements usually have to be performed
in ultra-high vacuum.
On the other hand, the tunnel current is sensitive to material composition and strain.
Atomic resolution in both lateral and vertical directions makes STM an ideal tool for
the investigation of growing surfaces and facets at this scale, which can give insight into
growth mechanisms. STM systems aflached fo a grow,h chamber allow measurement's
to be made without breaking the vacuum after growth.
The tunnel current ill STM is sensitive only to a thin layer al a sample surface, and
therefore it might seem that buried strucnLres are beyond the scope of STM studies.
However, buried structures can be studied by STM. Indeed, after cleaving samples STM
can be performed at the cleavage edge. Such a cross-sectional STM can reveal details
on the inner structure of buried nano-objects. In Fig. 5.9(a), the measured and simulated
cross-sectional STM profiles for a stack of InAs islands on GaAs substrate are shown. It is
seen that the buried InAs islands have the shape or truncated pyramids. A compositional
intermixing in the islands was found, with the GaAs composition decreasing linearly
from 0.4 at the base to 0 at the top of the islands. The corresponding lattice-parameter
distribution in the growth direction is shown in Fig. 5.9(c). This indicates directly an
increase of compressive strain in the GaAs matrix above and bclow the islands.
(b)
height ( ) 0 I 2 3 4
6
2
" -2 -2
-5 0 5 10 -10 o 10 -10 0 10
position along [OOI} (nm) .I'lllIIOJ(nm) YII (110] (nm)
Figure 5.9 Cross·sectional seaMing tunneling microscopy (STM): (a) an image of a stack of
(nAs islands in GaAs; (b) comparison between measured and simulated height profiles for a
similar sample; (e) the lattice parameter in the growth direction in an InAs island (the
experimental data were obtained from cross·seetional STM; the solid line is from a simulation
assuming an In conlent increasing from island base to island apex); and (d) the electronic
\\lUvefunction measured at two different tip biases. compared with simulations for the ground
and the first excited stales. Reprinted with permission from 1. Stangl, V I-Ioly er 01., "Structural
properties of self-organized semkonductor nanostructures," Figs. 25 and 26. Re\( Mod. Phys.
76,725-783 (2004).
II is remarkable that, apart from providing structural information, low-temperature

scanning tunneling spectroscopy has been used for wOIlefimcfioll mapping of single.
electron slates in nanostructures. In Section 3.4, while studying the electron states in
spherically symmetric potentials we introduced the states of different symmetries: s, p,
etc. When they are applied to the InAs dots (islands) the STM methods directly reveal
S·, p., d·, and even f·typc states as made visible by an asymmetry of the electronic
structure. which can be attributed 10 a shape asymmetry of the islands. Simulation of

the electron ground state and first excited state of an InAs island corresponds well
to the STM image, showing that the wave functions in such islands are indeed atolll-
like; see Fig. 5.9(d). In Ihe panel on the left, these are electrons in the ground Slate;
in the panel on the right, electrons from both Ihe ground state and the first excited
state contribute to the measured electron distribution. These Iwo measurements were
performed at different voltages at the STM tip: at a low bias of 0.69 'Y, only s electrons
contribute, whereas al a larger bias of 0.82 V. both sand p electrons contribute to the STM
Image.
Probe Laser
Four-quadrant
Photodetector
Agure5.10 Alomic·force microscopy in the contacllllode. The lip has dimensions ofabolll
30-50 nm al the end.
Atomic-force microscopy
An atomic-force microscope measures the force between the sample surface and a very
fine lip. The force is measured either by recording the bending of a cantilever on which
thc tip is mounted - the contact mode - or by measuring the change in rcsonance fre-
quency due to the forcc- the tapping mode. For example, in Fig. 5.10, the contact-mode
AFM technique is sketched. With a typical resolution of several nanometcrs laterally
•
and several A vertically, AFM is ideally suited to characterize the shapes ofnanostruc-
tures. For large scan sizes up to 100 ~m x 100 ~m, the lateral arrangement can also be
obtained. With AFM, any surface can be investigated; hardly any sample preparation is
required. A drawback of AFM is that only structures on a surface can be investigated.
Furthermore, most semiconductor materials oxidize under ambient conditions, so that,
strictly speaking, the AFM images usually show the surface of this oxide. When obtain-
ing quantitative data such as lateral sizes and heights of structures, this has to be kept in
miml as well as the fact that the image is actually a convolution of the samples surface
morphology with the shape of the microscope tip. Figurc 5.10 is a schcmatic view of a
contact-mode AFM. Essentially, a micrometer-size cantilever has an extremely sharp tip
auachcd 10 it, which is sharpened to about 30---50 nm at the cnd. A low-powcr probe laser
beam is rcflected off the top of the cantilever and into a four-quadrant photodetector,
which records the position of the reflectcd beam. Note that the probe beam need not be
perfectly aligned (as long as some part of the beam is reflected into the detector, and the
surface does not reflect too heavily into the detector), and need not even be smaller than
the detector (since the difference between the quadrant signals allows the determination
of the beam position). The photodetector measures the position of the reflected beam,
which in turn gives information abollt the position of the cantilever and hence the tip. If
the whole apparatus is raster-scanned across the surface (or the sample is scanned under
the microscope), then an image of the surface relief can be generated.
(b)
Figure 5.11 PbSe islands with [00 I]-type facets: (a) an AFM image orlhe lOp surface of a
PbSelPbEuTe island multilayer; (b) an AFM image oran area of) ~n x J ~llll oflhe lOp
surface of a PbSelPbEuTc isl:md multilayer. Islands are arranged in a regular array up to the
sixth·nearest neighbor. Reprinted with pcmlission, from J. Stangl, V. Holy et al.• "Structural
properties of self-organized semiconductor nanostructurcs;' Figs. 25 and 26, Re\( Mod. Phy.f.
76,725-783 (2004).
Examples of the quantitative analysis of AFM images arc shown in Fig. 5.11. There,
the lOp surface ofPbSe/PbEuTe multi layers is shown. Both materials are semiconductors.
From Fig. 5.11 (a), one can see thai PbSe forms triangular pyramids with [00 I)-type side
facets. The lateral ordering can also be analyzed by AFM. In Fig. 5.II(b), a hexagonal
in-plane arrangement of the pyramids is evident.
Transmission electron microscopy and scanning electron microscopy

Among the methods which allow one to "see" things at the nanometer scale, two types of
electron microscopy play an important role. Transmission electron microscopy (TEM)
makes possible the visualization of thin slices of material with nanometer resolution.
This technique has subnanometer resolution, and, in principle, can resolve the electron
densities of individual atoms. A TEM operates much like an optical microscope, but
uses electrons instead of visible light, since the wavelength of electrons is much smaller
than that of visible light. As we have already discussed, the resolution limitation of any
microscopy is based on the wavelength of the probe radiation. As studied in Chapter 2,
the electron wavelength is much smaller than that of visible light. Since electrons are used
instead of light. glass lenses are no longer suitable. Instead, a TEM uses magnetic lenses
to deflect electrons. Beyond this, a TEM is very similar to a conventional microscope,
complete with condenser lenses, objective lenses, and projector lenses.
In a TEM, the electrons are collimated from the source and passed through the sample,
and the resulting pattern of electron transmission and absorption is magnified onto a
viewing screen. The image is typically recorded with a charge-coupled-device (CCO)
camera, whose working element is a Si chip with the surface divided into a large array of
pixels sensitive to charge carried by electrons. In scanning electron microscopy (SEM),
Scanning Electron Transmission Electron

Microscope Microscope
Source
•• I
I condenser/
• I
•
Scan Coils I •
• Projector Lens
Screen
Detector
Figure 5.12 Schematic diagrams of a scanning electron microscope (SEM) and a transmission
electron microscope (TEM).
the electron beam is not projected through the whole sample area. Instead, it is rastcr-
scanned across the surface, and the secondary electrons, or X-rays, emi"ed from the
surface are recorded. This generates a lower-resolution image, but allows the direct
mapping of surface features, and can even be used for elemenlal analysis. Both types of
electron microscope are depicted schematically in Fig. 5.12.
The electron microscopy techniques are widely used due to their very high spatial
resolution and sensitivity to composition. TEM can be performed either on thin slices
parallel to the sample surface (plane-view TEM) or on cross-sectional slices. Hence,
buried nanostructures can be examined well by TEM, with some restrictions due to
specimen preparation: in many cases, the latcral island diameter is comparable to the
slice thickness. The image analysis is often not straightforward but requires elaborate
image-analysis teclmiques and/or model calculations. Compared with other techniques,
usually very small areas are investigated, so that no statistically averaged values can be
obtained.
We present here only one example of atomically resolved transmission electron
microscopy, from which the positions of unit cells, strain, and composition informa-
tion were derived. The results obtained, after digital analysis of the lattice image, are
given in Fig. 5.13: the strain distribution (i.e., distortion of atoms constituting the island
from their regular positions) for an InGaAs island on a GaAs substrate is depicted.
Remarkably, these techniques make it possible to visualize a detailed map of the strain
for an object of size a few tens of nanometers. From Fig. 5.13. it is seen clearly how
5.5 Spontaneous formation and ordering 127
111 m
~ ,
Figure 5.13 Strain distributions obtained rrom the TEM images or InGaAs islands in GaAs by
using the method of digital analysis oflatticc images. Reprinted with permission, from J. Stangl.
V. Holy er al., "Structuml properties ofself-organized semiconductor nanostructures," Fig. 27(a).
Rev. Mod. Phy.,. 76,725-783 (2004).
the strain increases at the apex of the InGaAs island, while around the island the strain
changes its sign.
In conclusion, powerful characterization techniques have been developed to study
nanosize objects. The techniques give three-dimensional images in real space and on
an alomic scale in all three dimensions. The methods are nondestructive. They provide
means to perform stnlctural and chemical analyses of the materials used in nanostruc-
tures. Moreover, these techniques make it possible to observe and measure directly the
electron distributions inside the nanostructures; that is, it is possible to observe the
electron probability densities!
5.5 Spontaneous formation and ordering of nanostruclures
From the previous section, it is evident that crystal-growth and device-fabrication tech-
niques arc highly developed and are already having an impact on nanoscale semiconduc-
tor structures and devices. Every step toward realizing and perfecting artificial nanos-
tructures involves a nwnber of new physical and chemical processes and requires very
serious efforts and technical innovations.
Importantly, Mother Nature point's to another way 10 produce nanostruclures. Indeed,
the phenomenon of spontaneous formation of periodic domain structures in solids with
a macroscopic periodicity has been known for several decades. Progress in TEM, STM,
and AFM is facilitating the reliable and accurate observation, investigation. and control
of surfaces and periodic slructures with the characteristic periodicity of 1-100 nm. This
opens a new way to usc self-organizing growth processes as well as the formation of
periodically ordered structures on semiconductor surfaces for the direct fabrication of
quantum nanostruchtres and devices.
d d
(a) (b) (e)

Flgure5.14 Three types of periodic structures: (a) periodically faceted surfaces; (b) planar
domains; and (c) three-dimensional strained islands.
[001]
L[IIO]
Unreconstructed Reconstructed
Figure 5.15 Reconstruction of Si surface.
There arc three distinctive classes of spontaneously fOfmed nanostructures all a sur-
face, as illustrated in Fig. 5.14. These nallostfuctures arc (I) periodically faceted surfaces
(Fig. 5.14(a», (2) periodic structures of planar domains (Fig. 5.14(b», and (3) ordered
arrays ofthrec·dimensional coherently strained islands in lauice-mismatched hClcroepi-
taxial systems (Fig. 5.14(c». Despite the facl that the geometries of the three classes
are different, there exist common features for all of these nanostructures. The main one
is that the driving force of the periodic ordering is a long-range clastic interaction. The
elastic·strain field is created due to the discontinuity of the intrinsic elastic properties on
domain boundaries ancVor by the lattice mismatch between two materials composing the
heteroepitaxial system. The three classes depicted in Figs. 5.l4(a)-(c) arc equilibrium
structures. In particular, they can be observed upon annealing of the crystal, or upon
interruption of the crystaJ growth. Here we restrict our discussion to the formation of
three-dimensional coherently strained islands.
We begin with a brief discussion of the strain arising at surfaces of solids. Since atoms
in the surface layer of any material are in a different environment from that experienced
by those in the bulk, the surface layer energetically favors a lattice parameter different
from the bulk value in the directions parallel to the surface. Being adjusted to the bulk
lattice, the surface layer is intrinsically stretched or compressed. Therefore, the surface is
characterized by intrinsic surface stress. At the surface, even the symmetry of the crystal
can be changed; this is known as sUI/ace recollstruction. Figure 5.15 illustrates such a
reconstruction of the Si surface.
5.5 Spontaneous fonnation and ordering 129
FM vw SK
Figure 5.16 Three growth modes ofhclcroepilaxial systems: Fmnk-van def Menvc (FM),
Volmer-Weber (VW), and Slnmski-Kraslanow (SK).
To some extent the intrinsic surface stress ofa solid is analogous to the surface tension
of a liquid. However, there is a fundamental difference between the properties of a )jquid
surface and those of a solid surface. Indeed, liquids are typically isotropic. The processes
of formation and deformation of a liquid surface are identical and can be described by
a single parameter that characterizes the energy of cutting of bonds on the surface.
This is the so-called slIIface energy. However, in crystals the process of the formation
of a surface is quile different: the distance beh.\'een atoms changes and symmetry can
change. This is quite different from the simple cutting of bonds. In addition, crystals are
anisotropic and the energy needed to create a free surface of a given orientation depends
on the orientation.
These features of the surfaces of solids give rise to different epitaxial growth regimes.
Three growth modes are distinguished traditionally. They are the (I) Frank-van der
MelWe (FM), (2) Volmer-Weber (VW), and (3) Stranski-Krastanow (SK) growth modes.
Respectively, they can be described as (I) layer-by-Iayer growth, (2) island growth (three-
dimensional), and (3) layer-by-Iayer plus island growth. 111ese three modes are illustrated
in Fig. 5.16. The particular growth mode for a given system depends on the surface
energies and on the lattice mismatch between the material of the substrate and that of
the grown layer. The growth regimes discussed in Section 5.2 are of the FM type.
In the following discussion we define the surface energy as the e.xcess energy ofa ve,y
thi" distorted layer (one or two monolayers) at the free surface ofa solid. The interface
energy is the excess energy of a thin layer at the interface between two solids. In lattice·
matched systems, the growth mode is governed only by interface and surface energies.
Let YI and Yl be the surface energies of the substrate and epitaxial layer (epilayer),
respectively, and let YIZ be the interface energy. Then, if the sum of the epilayer surface
energy and the interface energy is lower than the energy of the substrate surface, i.e.,
YI > Yl + YIl, the deposited material wets the substrate and the FM growth mode occurs.
A change in yz + YIZ can drive a transition from the FM to the VW growth mode. These
two modes ofcoexistence of crystalline materials are quite analogous to those of a liquid.
For a strained epilayer with small interface energy, YIZ. the initial growth may occur
layer.by.laycr; however, as the layer becomes thicker it may lower its increasing strain
UF
DI
CI
Figure 5.17 Elastic-Slrain relaxation during Stranski-Krastanow growth (schematic). Light gray
areas denote the substrate, dark gmy areas denote the lanice-mismatched epilayer. The lines
symbolize lattice planes. Top, uniformly stmined film (UF); middle. dislocated relaxed islands
(01); and bottom. coherently stmincd island.. (el).
energy by forming isolaled islands in which strain is relaxed. This results in the SK
growth mode. Thus. the SK mode depends criticalJy on the lattice mismatch. Three
scenarios for strain relaxation arc sketched in Fig. 5.17. Figure 5.17(a) corresponds to
uniform pseudomorphic growth without any stress relaxation, which can ex.ist for a very
thin strained layer, as discussed in Section 4.7. Figure 5. I7(b) corresponds to a dislocated
relaxed island. Finally, Ihe island in Fig. 5.17(c) is coherently strained. From the point
of view of the use of self-forming nanosize islands. coherenlly strained structures are of
most importance, because of their high quality.
The scenarios presented in Fig. 5.17 occur as a result of the interplay of several
parameters: (I) the ratio of the surface energy and the energy of the dislocated interface,
(2) the amount of deposited material Q, and (3) the lattice mismatch, € (see Eq. (4.23)
and Fig. 4.16). As an example. a diagram of possible growth regimes is depicted in
Fig. 5.18. In Fig. 5.18, the various growth regimes are identified in terms of the amounl
of deposited material, Q, versus lattice mismatch, €.
For a dense system of islands, the clastic interaction between islands via deformation
of the substrate is essential The system of interacting islands is then a system of elastic
domains where the energy minimum corresponds to a structure with periodic domains.
Thus. there is a possibility of growing an ordered system of nano-islands.
Following this general analysis of formation of three-dimensional islands under pseu-
domorphic growth of crystalline materials, we consider particular examples of such
self-growing nanosize helerostructures.
The first experimental evidence of the formation of the coherently strained islands
was obmined by transmission electron microscopy of the InAs/GaAs system (GaAs is
the substrate, while InAs is the epilayer). As follows from the data of Table 4.8, Ihis is
the lattice·mismatched system with mismatch parameter € ~ 7%. Coherently strained
and stable islaJlds have been found for many systems: Ge/Si, GeSi/Si (Si substrate of
[100] orientation), AllnAsfGalnAs. InAslInP, CdSe/ZnSe, and others.
5.5 Spontaneous formation and ordering 131
R,
" " "

5 r--..-+--f----'f----,
R, R,
4
3
012
FM
SK,
I
vw
o'-;;--~-----,!-:--~--;!
0.0 0.1 0.2
,
FM SK vw
Figure 5.18 An equilibrium phase diagram of a lauice-mismatchl:d hclcroepitaxial system as a

function or the total amowlt of depositcd material Q and the lattice mismatch E. The small
panels at the lOp and bottom illustrate the morphology orlhe surface in the six growth modes
described in the lexl. The small empty triangles indicate the presence of stable islands, while the
large shaded ones refer to ripened islands. Reprinted with pennission from I. Daruka and A.
Barabasi. "Dislocation-free island formation in hclcrocpitaxial growth: a study at equilibrium;'
Phys. ReI'. Leu., 79, 3708 (1997). ttl 1997 by the American Physical Society
To study the formation of the islands, the growth-interruption method typically is

applied. This method allows one to control the amount of the deposited material even
below one deposited monolayer at the same growth temperature. For example, at a growth
temperature of 4S0 °C the following features were revealed. When the average thick-
ness of InAs deposition reaches a critical value of 1.6-1.7 monolayers, a morphological
transition from an InAs layer to threc-dimensionallnAs islands occurs. Aftertwo mono-
layers of InAs have been deposited, an array of well-developed islands is observed. The
average lateral size, the height, and the density of the islands arc 100-140 ,50-SO A,
and 10 1°_lOll cm- 1 , respectively. Figure 5.19 depicts a single lnAs island on GaAs
substrate studied by scanning tunneling microscopy.
In general, results of growth and post-growth annealing studies show that the islands
grown in the SK mode are usually stable against ripening. Under given growth conditions
they have well-defined sizes and shapes. For example, in SiGe/Si, essentially only four
forms of islands are observed: (I) shallow pre-pyramids, (2) square pyramids, (3) "hut
clusters" - elongated pyramids, and (4) large domes with facets in several directions.
Figure 5.20 depicts a diagram that links possible shapes of islands with their volume
and the mismatch parameters for Ge or GeSi on Si substrate. In the first stage of growth.
shallow pre-pyramids appear that later convert to pyramids and then to domes. Pyramids
and domes are clearly observed during growth at higher temperatures, whereas the much
smaller hut clusters fonn at lower growth temperatures. The typical lateral sizes of the
132 Growth t fabrication. and measurement techniques
[110]
[110]
...
50A
Agure 5.19 A single InAs island on GaAs substrate. Reprinted with permission. from 1. Stangl. V.
Holy el aJ.• "Structural properties of self-organized semiconductor nanostructures:' Fig. 12(b),
R,,,,. AI",I. Ph)'s. 76.725-783 (2004).
1.0
4.0 ~
0.8 ~
~
domes
u
0.6
3.0
-"'"
.<:
E
" 2.0 .-
~
'" 0.4 ".-"

0.2 1.0 -"
'"
..J
0.0 0.0
10' 10' 10' 10' 10' 10'
Island volume (nm 3)
Rgure5.20 Shape lrnnsition orGe orSiGc islands grown on Si [001) during growth (solid arrow).
postgrowth annealing (dotted arrow). and Si capping (dashed arrow). The solid curves represent
the critical volumes for pyraJluds and domes. Corresponding STM images arc embedded in the
diagrJl1l. Reprinted with permission. from 1. Stangl. V. Holy el aJ.• "Structuml properties of
self-organizcd semiconductor nanoslructun:s:' Fig. 14. Re\t MOll. Phys. 76. 725-783 (2004).
• •
islands arc 800--1,000 A and their heights are 60-100 A. All of these features of island
growth can be explained by analyzing surface and interface energies, as \vcll as the
elastic energy of strained materials. Crystalline anisotropy is also an important factor in
the fonnation of islands.
As for lateral correlation of the island. the mutual influence of islands on the same
surface is rather weak and can be observed experimentally only for growth at very
low rates; i.e.. close to thermodynamic equilibriwu. Typically. the lateral self-ordering
5.5 Spontaneous fonnation and ordering 133
nm
mo
00 nm
520
260·f~.'
o
~lm
o0
Figure 5.21 AFM images of the surface ora Ge layer grown on lithographically prcpaucmcd Si
[001] substrates. In the sample shown on the lOp left. the islands are arranged in a regular array
along two orthogonal [II OJ-directions. In the sample shown on the bottom right, the unit vectors
of the two-dimensional array of pits arc oriented along the (100]· and [I IO]-directions. leading
10 a 45 0 island alignment. X-ray diffraction (lOp right and hallom left) demonstrates the high
degree of ordering of islands. Reprinted with permission from J. Stangl, v: Holy ef 01..
"Structural properties of self-organized semiconductor nanostructurcs." Fig. 21. Rev. Mot/.
Phy.,. 76,725-783 (2004).
of islands can yield short-range ordered patterns with astonishing regularity; however,
this process does not result in true long-range ordering. Ordering can be realized by
using island nucleation on lithographically prepauerned substrates. Similarly to the case
for nonpatlcrned substnHes, for prepatterned substrates the nucleation of the islands
is affected by the surface curvature and the surface stress. Thus, the substrate can be
prepared to provide initial nucleation at selective locations on the substrate. Various
methods can be used; e.g. shallow etching of trenches in Si and subsequent overgrowth
with a GeSi multilayer, followed by Gc nucleation; and direct growth of Ge in two-
dimensional periodically etched pits. The latter method results in long-rangc-ordered
island structures, as shown in Fig. 5.21. Nucleation occurs at the intersection of the side
facets within the pits. Thus, a combination of self-organizing processes in crystal growth
with prepatterning methods may be used to achieve precise placement of individual
islands, as well as high ordering throughout nanosize heterostructures.
In conclusion, for device applications, uniform islands with controlled positioning are
required. This goal is di fficult to achieve using self-organization alone, but a combinat ion
with other techniques may be successful. The following three routes arc most promising:
(I) a combination with conventional lithography will allow controlled positioning ofsclf-
organizing nanostructures without losing their inherent advantages; (2) seeded growth
with catalytic nanoparticles facil itates the fabrication of structures significantly smaller
than the SK islands; and (3) a combination of self-assembly with the techniques of
organic chemistry and biochemistry. Nowadays, the controlled fabrication of extremely
small structuresofa few nanometers in all three dimensions and their selective addressing
seems quite feasible.
5.6 Clusters and nanocrystals
The spontaneously formed and ordered nanostructures studied in the previous section
are merely one type of a number of known examples of nanoclusters and nanocrystals,
which can be grown by using various technological methods.
A nanoclustcr can be thought of as a size.dependent collection of atoms (from several
atoms to several thousands of atoms). In an ideal case, the cluster is isolated. Isola·
tion means the absence of "foreign" chemical species within the cluster volume, or on
its surface. Obviously. the simplest way to realize isolation is to synthesize a cluster
under vacuum conditions and then keep it in an inert-gas environment. Typically, an
ideal cluster has a high density of unsatisfied dangling bonds on its surface. Such ideal
unsupported clusters are not very useful for fimcfional nanostructured materials and
devices. Indeed, one needs to manipulate the clusters, to place them onto a surface in
a certain order, to provide interaction with them, etc. However, if the total number of
atoms in a c1uster;V is large, the fraction of atoms at the cluster surface which can have
unsatisfied bonds is of the order of;VI. The ratio of atoms on the surface to N decreases
as lIN! and a big cluster is almost an "ideal" cluster. Importantly. the dangling bonds
have a high reactivity and thus a semiconductor cluster prepared under high·vacuum
conditions will readily be oxidized on exposure to the atmosphere. Practically, cluster
dangling bonds can be terminated artificially by using an organic, or inorganic, addi-
tive. Such a passivation ofa cluster surface leads to an effective functionalization of the
cluster.
The following two techniques of cluster and nanocrystal fabrication are different from
the growth methods considered in the previous section: gas-phase and colloidal cluster
synlheses.
Under gas·phase cluster synthesis, clusters are grown in a gas, prior to their passivation
and deposition onto a surface. Such clusters are formed when the vapor pressure of
atoms composing the clusters is much larger than it should be under equilibrium at a
given temperature. Nonequilibrium atom vapors can be created by various methods, for
example by laser evaporation of a solid, laser- or thermally induced decomposition of
species conlaining necessary atoms, elc.
For example, vacuum laser-induced decomposition of silane leads 10 formation of Si
clusters as ultra·small nanocrystals. Typically, the nanocrystal sizes nmge from 3 nm to
10 nm. The clusters can be deposited onto the surface ofa metal, graphite. or silicon. In
Fig. 5.22, an STM image of Si clusters is presented. One can see the atomic structure
of the Si-(III) surface and sevcral Si clusters of various sizes. On the left of the image.
the detailed structure ofa larger Si cluster is observable. Interestingly, the surface of the
Si crystalline substrate oriented normally to the (III) direction has a large reactivity to
5,6 Clusters and nanocrystals 135
Rgure 5.22 Silicon nanoclustcrs absorbed on a (III) surface ofSi. Several clusters, ofdifTering
sizes, are seen on this STM image. Reprinted with permission from D. W. McComb, B. Collings
el al. "An atollHesolved view of silicon nanoclusters." Chefl/. Phys. Leu. 2S I, 8-12 (1996).
Pemlission from Elsevier.
Si cluslers. This explains the large sticking coefficient for 5i clusters. As a result, on
being deposited onlo the surface, Ihese clusters do not diffuse onto the substrate at room
temperature.
Compound semiconductors can also be synthesized in the gas phase. For example.
laser evaporation of materials containing Ga and As atoms in an inert gas produces
GaAs clusters with dimensions of tens of nanometers. If these clusters arc exposed to
air, they are capped by an As-rich oxide shell. Then, the clusters can be deposited
onto a surface (Si-(IOO). for instance) and manipulated by using an atomic-force
Illlcroscope.
The colloidal synthesis of almost monodisperse nanoparticles is based on controlled
nucleation and growth of clusters in a precursor-containing solution. Both metallic and
semiconducting clusters can be grown. The most studied case is synthesis of III-V
and II-VI compounds. The method involves injection of reagents containing the cluster
constituents (for example, Cd and Se) into a hot solvent, where nucleation of CdSe
occurs. Molecules of the solvent cap and thereby passivate the nucleated cluster. For
the above example. l,.i-,,-oc~\llp"osphide (TO PO) is used as a solvent. The reagents are
Me,Cd and TOPSe.
A hot solution of TOPO is favorable for nucleation of Cd e. TOPO capping the
clusters significantly slows the cluster growth and makes the properties of the product
more controllable. Careful control of the solution temperature enables production of
nanoparticles with a small size dispersion. The size dispersion of nanocrystals can be
within a few percent of the average diameter. The latter is a few to tens of nanome·
ters. The shape of nanocrystals is close to spherical. On being deposited onto a surface.
these nanocrystals form an ordered lateral structure. In Fig. 5.23, a tunneling elec·
tron microscope image is presented, which provides evidence of the formation of a
Rgure 5.23 A transmission electron microscopy image ofCdSe n:mocrystals of size 4.8 nm
grown by the colloidal method and then deposited on a surface. X-my diffmction demonstnlles
the high ordering of these nanocrystals. Reprinted with permission, from P. Moriarty,
"Nanoslrueturcd materials;' Fig. 19(b), Rep. Prog. Phys. 64. 324 (2001). CO lOP Publishing
Limited.
sel f-organized 1O-nm CdSe nanocrystal superlattice. Lnterestingly, artificial passivation

ofthese clusters modifies intercJuster interaction. The latter is responsible for intercluster
spacing. Thus, the ordering ofedSe and other clusters on a surface can be controlled by
the regime of their fabrication in the colloid. In general, colloidal chemistry is conve-
nient for "engineering" of basic properties of nanocrystals. including their shape, size.
and state of surface. For example, more sophisticated nanocrystals prepared in a colloid
can include a multilayered coating: the CdSe clusters can be covered by a layer of HgS
followed by a layer of ZnSe and so 011.
Clusters and nanocrystals possess a number of properties different from those of
bulk materials. Indeed, they occupy an intermediate place between individual natural
atoms/molecules and bulk crystals. For semiconducting materials, the clusters have
electron energy bands practically the same as those of corresponding bulk materials.
However, their small sizes restrict electron motion, giving rise to electron confinement
and quantization in all three directions. As a result. the fundamental electrical, optical.
and mechanical properties are modified. In the next chapter we will study some of these
modifications.
5.7 Methods of nanotube growth
From the discussion of carbon nanotubes in Chapter 4, it is clear that the properties
of these nano-objects are considerably different from the properties of other nanos-
tructures and nanodevices fabricated on the basis of bulk-like materials. The same
is valid for methods of growth of nanotubes. In this section, we consider several
growth methods for carbon nanotubes. Basically, these methods are arc-discharge, lascr-
ablation. and chemical-vapor deposition. Arc-discharge and laser-ablation methods were
5.7 Methods 01 nanotube growth 137
Agure 5.24 Catalytic growth of a (10, 10) armchair carbon nanotubc using a metal atom (large
black sphere) as a catalyst. Carbon aloms arc depicted by while balls. Reprinted with
permission. from 1. Charlier and S. lijima, "Growth mechanisms:' in M. S. Drcssclhaus,
G. Drcsselhaus, and Ph. Avouris (Eds.), Carbon Nanowbes. Fig. II, Topics in Applied Physics
vol. 80 (Berlin. Springer-Verlag. 2001), 55-79.
historically the first approaches used to fabricate nanotubcs. Chemical-vapor-deposition

growth methods arc new and highly promising for scale-up of defect-free nanotubc
materials.
Carhon nanotubes are constituted solely of carbon atoms. Thus, any growth method
for these nanoslruclures should first provide carbon atoms, and then condense the car-
bon vapor under certain nonequilibrium conditions at temperatures below the graphene
mclting point ("'4100 K).
Carbon-vapor condensation can result in several different forms of carbon-containing
products: fullerenes, graphitic layers, nanotubes, etc. To obtain a considerable yield of the
nanotubes, corresponding growth regimes should be used. For example, relatively high
temperatures (1000-1300 K) are necessary to form single-walled nanotubes. Indeed,
the rolling-up of a graphene sheet into a tube requires additional energy, especially for
small diameters of single-walled tubes. Then, it was found that the key role in nanotube
growth can be played by catalytic processes involving transition metals (iron, nickel,
cobalt, etc.). Particular catalytic mechanisms are complicated.
Figure 5.24 illustrates one such mechanism for growing a (10, 10) armchair nanotube
(white balls) with a Ni (or Co) atom (large black sphere) chemisorbed onto the open tube
edge. Though Co and Ni atoms are strongly bound, they are still very mobile al the edge
of the growing tube. Consequently, the metal catalyst keeps the tube open as a result
of it'S mobility around the open edge, ensuring that any pentagons or Dlher high-energy
local structures are rearranged to hexagons. The latter occurs through the exchange
mechanism: the metal catalyst assists two incoming carbon atoms (or a C 2 molecule) in
the formation of carbon hexagons, thus increasing the tube's length.
(a)
He
Arc-discharge J
(b)
Laser ablation
Laser .-t::t:==~::!!!!!!!!!!!!!!!!...._--4-.J
pulse
Oven temperature -1200 °C
(c)
CVD
Oven temperalure -500-1000 °C
FIgure 5.25 Schematic diagrams of arc-discharge (a); laser ablation (b); and chemical-vapor-
deposition (c) methods of growth of carbon nanotubes.
Arc-discharge and laser ablation

In an arc-discharge, carbon atoms are evaporated by a plasma of helium gas ignited
by high currents passed through opposing carbon anode and cathode as illustrated by
Fig. 5.25(a). Evaporation ofcarbon atoms in an arc-discharge isa strongly nonequilibrium
process, thus these atoms should undergo sequential condensation at temperatures below
the melting point of graphite (3000 K). Carbon nanotubes arise as one of the products
of this process. The synthesized multi-walled nauorubcs have lengths of the order of
10 ~ and diameters in the range 5--30 nm. The nanotubes are typically bound together
by molecular van der Waals interactions and form tight bundles.
For the growth of single-walled tubes, a metal catalyst is needed in the arc-discharge
system. For example, producing substantial amounts of single-walled nanotubes is pos-
sible by arc-discharge with the use of a carbon anode containing a small percentage
of cobalt catalyst in the discharge camera. As a result, abundant single-walled nan-
olubcs are generated in the soot material. Optimization of the growth of single-walled
carbon nanotubcs in an arc-discharge is achieved by using a carbon anode contain-
ing a large atomjc percentage of a transition metal (for example, up to 4% nickel
catalyst).
5.7 Methods of nanotube growth 139
Agure 5.26 A bundle of carbon nanotubcs grO\vn by laser abl:lIion. Reprinted with permission,
from Hongjic Dai, "Nanotubc growth and characterization;' in M. S. Dressclhaus,
G. Dressclhaus. and Ph. Avouris (Eds.). Carbon Nunotubes. Fig. 4, Topics in Applied Physics.
vol. 80, pp. 29-53 (Berlin, Springer-Verlag, 2001).
The process ofgrowth of high-quality single-walled nanolUbes takes place also during
the so-called laser-ablarion (laser-oven) method illustrated by Fig. 5.25(b). The method
utilizes intense laser pulses to ablate a carbon target containing 0.5 atomic percent of
catalytic nickel and cobalt. The target is placed in a tube-furnace healed to 1200 o e.
During laser ablation, a flow of inert gas is passed through the growth chamber to
carry the grown nanohlbcs downstream to be collected on a cold finger, where carbon
condensation occurs. The resulting tubes are primarily in the form of ropes consisting
of tens of individual nanotubes close-packed into hexagonal crystals via the van der
Waals interactions. Packaging of individual nanotubes into a bundle is clearly seen in
Fig. 5.26.
Typically, arc-discharge and laser-ablation methods lead to a number of by-products:
fullerenes (graphitic polyhedrons, some with enclosed metal particles), amorphous car-
bon, etc. Further purification is necessary to obtain the nanotubes. This process involves
refluxing the nanotubes in a nitric acid solution for an extended period ohime, which oxi-
dizes away amorpholls carbon particles and removes metal catalyst species. Production
of single-walled nanotubes of high quality by laser ablation and arc-discharge is rcla-
tively simple and has already led to the wide availability of samples useful for studying
their fundamental properties.
Chemical-vapor deposition
A schematic experimental setup for chemical-vapor-deposition growth is depicted in
Fig. S.2S(c). The growth process involves heating a catalyst material to high temper-
atures in a tube furnace and flowing a hydrocarbon gas through the tube reactor for a
period oftime. Materials grown over the catalyst are collected upon cooling the system to
room temperature. Parameters controlling nanotubc growth are the hydrocarbons, cata-
lysts, and growth temperature. The active catalytic species are typically transition-metal
nanoparticles formed on a support material such as alumina. The growth mechanism
involves the dissociation of hydrocarbon molecules catalyzed by the transition metal,
and dissolution and saturation of carbon atoms in the metal nanoparticle. The precip-
itation of carbon from lhe saturated metal particle leads to the formation of hlhular
carhon solids. Tubule formation is favored over formation of other forms of carhon
such as graphitic sheets with open edges. This is because a tube contains no dangling
bonds.
Relatively high temperatures (1000-1300 K) are necessary to form single-walled
nanotubes with small diameters and allow the production of nearly defect-free nanotube
structures. Among all hydrocarbon molecules, methane is the most stable at high tem-
peratures against self-decomposition. The methane chemical-vapor-deposition approach
is promising for enabling scale-up of defect-free nanotubc materials to the kilogram or
even ton level.
Directed growth of single-walled nanotubes

Ordered, single-walled nanotube structures can be grown directly by methane chemical-
vapor deposition on catalytically patterned substrates. Consider, for example, a method
developed to grow suspended nanotube networks on substrates containing lithograph-
ically patterned silicon pillars. The growth starts with developiJlg a liquid-phase cat-
alyst precursor material that has the advantage over solid-state catalysts of allowing
the formation of uniform catalyst layers for large-scale catalytic patterning on surfaces.
The precursor material consists of a triblock copolymer, aluminum, iron and molybde-
num chlorides in mixed ethanol and butanol solvents. The aluminum chloride provides
an oxide framework when oxidized by hydrolysis and calcination in air. The triblock
copolymer directs the structure of the oxide framework and leads to a porous cata-
lyst structure upon calcination. The iron chloride also can lead to catalytic particles
needed for the growth of nanotubes. The catalyst precursor material is first spun into
a thin film on a polydimethyl siloxane stamp, followed by contact printing to transfer
the catalyst precursor selectively onto the tops of pillars pre-fabricated on a silicon sub-
strate. The stamped substrate is calcined and then used in chemical-vapor-deposition
growth.
Remarkably, the nanotubes grown from the pillar tops tend to be directed from pillar to
pillar. The directed growth ofsuspended single-walled nanotubes is presented in Fig. 5.27
for three different configurations of the pillars: (a) a nanotube power-line-like structure,
(b) a square ofnanotubes, and (c) an extensive network of suspended nanotubes. Such
a directed growth can be understood as follows. Nanotubes are nucleated only on the
tower-tops since the catalytic stamping method does not place any catalyst materials on
the substrate below. As the nanotubes lengthen, the methane flow keeps the nanotubes
floating and "waving in the wind" since the flow velocity ncar the bottom surface is
substantially lower than that at the level of the "lower-tops." This prevents the nanohlbes
from being caught by the bottom surface. The nearby towers on the other hand provide
5.8 Chemical and biological methods 141
Figure 5.27 Pillar tops are connected by suspended single-walled nanotubcs that form (a) a
nanotube power-line-like structure; (b) a square of nanotubcs; and (c) an extensive network of
suspended nanOlubes. Reprinted with pemlission, from Hongjic Dai." anotube growth and
characterization," in M. S. Dressclhaus. G. Dressclhaus, and Ph. Avouris (Eds.). Carboll
Nallolllbes, Fig. 9. Topics in Applied Physics. vol. 80 (Berlin. Springer- Verlag, 200 I), pp. 29-53.
fixation points for the growing tubes. If a waving tube contacts an adjacent tower, the
tube-tower van der Waals interactions will catch the nanotube and hold it aloft. By such a
growth mode the suspended nanotubes can be made of macroscopic length. Forexample,
tubes longer than 150 IJ.m can be grown.
Growth of isolated nanotubes on a specific surface site

It is of importance that the chemical-vapor-dcposition method allows one to grow
individual nanotubes at specific sitcs on flat SiOz substrates. The approach involves
mcthane chemical-vapor deposition onto substrates containing catalyst islands patterncd
by electron-beam lithography. Nanotubc chips witb isolated nanotubes grown from the
islands have already been obtained. This growth approach readily leads to tubes orig-
inating from well-controlled surface sites, and makes possible the development of a
controlled method to integrate nanotubes into addressable structures.
5.8 Chemical and biological methods for nanoscale fabrication
The technologies studied in previous sections are based primarily on physical processes
and methods. The use of chemical and biological methods opens new ways for nanofab-
rication. It is obvious that chemical synthesis can produce a wide range ofnanoparticles.
including inorganic, organic, and biological nanosystems. In addition, chemistry and
biology can be complementary to existing physical llano fabrication methods.
In particular, selected techniques of chemistry and biology portend alternatives to

conventional photolithography. One approach is based on self-assembly phenomena. It
is known that there afC molecules that, due to physical, biological, or chemical processes,
tend to congregate into long-range periodic structures. They can facilitate the realization
of reguJar periodic masks on the surface of a metal or a semiconductor with a character-
istic period of 10-5011111 and with internal pores ofdiameter 5-25 nm. Such resolution is
superior to that of conventional photolithography, whkh is now of the order of 100 11m.
Ln addition, the self-assembly methodology has potential for being mllch less expensive
and less time-consuming than the electron-beam lithography techniques currently used
for writing such small structures.
The chemistry methods can be used to tailor the chemical composition and structure
of a surface. This can be achieved through a new direct-write tool- dip-pen nanolithog-
raphy. which generates surface-patterned chemical functionality with a length scale of
1-100 nm.
In turn, achievements in biological science generate principally new methodologies
for preparing nanostructured materials with predefined and synthetically programmable
properties from given inorganic building blocks.
Chemical and biological innovations for nanotechnologies are briefly considered in
this section.
Chemical self-assembly of nanoscale structures

Here we consider chemical systems based on block copolymers. They exhibit self-
ordering, which can be used for nanopatterning.
A block copolymer is a macromolecule that consists of several polymer blocks, which
typically can be grown in a few steps. In Fig. 5.28 an example of a triblock po(vmer
is presented. The synthesis of this macromolecule starts with the polymerization of the
styrene monomer. The amount of monomer supplied for the first step is only sufficient
to create a mi"iaillre styrene chain (the first block) with an average degree of polymer-
ization of 9. (Note: the degree of polymerization represents the number of monomeric
units in a macromolecule or a block.) A second block with a similar average degree of
polymerization grows when another chemical - isoprene - is added. In the third step,
carbon dioxide (C0 2) installs carbo.\yl end grollps al the top of the previous miniature
diblock copolymers. One terminus of the triblock (the lower) is hydrophobic while the
opposite terminus is hydrophilic. These triblock molecules have a rod-eoil architecture
with a stiffrodlike segment (the second block) covalently connected to more torsionally
flexible segments. Chemical processes, typically, lead to a polydispersity of resulting
triblock molecules; however, their dispersity is small: the ratio of the mass-averaged to
number-averaged molecular masses ranges from 1.06 to 1.1.
Several triblock polymers can compose different clusters, as illustrated in Fig. 5.28.
On being deposited onto a surface. triblock polymers form a macromolecular thin film
with a high degree of self-ordering, as can be seen from the scheme in Fig. 5.29. The
triblock polymer film has a typical "standing-up" structure. The shaded circle indicates
a chemisorbing headgroup and the open circle an endgroup, which can be chosen from
FIgure 5.28 Molecular graphics of the triblock self·asscmbling molecule (leA) and a cluster of
triblock macromolecules (center and right). shade-coded for energy in the various seeton; of the
cluster. Reprinted with pennission from S., Stupp. V. LeBonhcur el aJ.. "Supramolccular
materials: sclf-organized nanoslruclures:" Science. 276. 384 (1991). ~ 1996 AAAS.
a variety of chemical functionalities. A top view of a real film obtained by transmission

electron microscopy is depicted in Fig. 5.30. The evident ordering of macromolecules
in the film is driven by the stiff rodlike segments of the triblocks. Specifically, a balance
of attrnctive and repulsive forces mediates the formation of macromolecular cells, as
for a crystal. In general, block copolymers easily cover the surfaces of various metals
and semiconductor crystals and form two-dimensional pcriodjc structures of various
symmctries ranging from low·symmctry to square and hcxagonal structures. Artificially
built polymcr films have thicknesses ranging from tens to hundreds of nanometers with
porcs of various sizes and various shapes.
144 Growth! fabrication. and measurement techniques
_End group
j . - - Backbone
• Head group
; - - Substrate
Agure 5.29 A schematic diagram ofa triblock polymer film: the polymer is shown "standing-up:'
After F. Schreiber, "Self-assembled monolayers: from simple model systems to
biofunctionalized interfaces:' J. Phys.: Condens. MOiler, 16, R881 (2004). Q lOP Publishing
Limited.
Figure 5.30 A transmission electron micrograph of a film formed by the triblock molecules,
revealing regularly sized and shaped aggregates that self-organize into superlattiee domains.
Reprinted with permission from S. Stupp, V. LeBonheur el tli.• "Supramolecular materials:
sclf--organizcd nanostructures:' Sciellce, 276. 384 (1997). Q 1997 AAAS.
TIlere arc two ways to create nanostructures using self-assembly of block copolymers:
(i) precipitation of metal or other inorganic crystals onto a polymer mask attached to a
substrate; and (ii) using the polymer mask for the subsequent processing of the substrate,
or as a support for additional auxiliary masks. After the mask has been formed on the
surface of a semiconductor, the device-fabrication technology remains the same as with
the conventional nanofabrication and employs dry ion or selective etching, oxidation,
metallization, diffusion, selective gro\\1h, etc.
Biological methods
As examples of the use of biological methods as elements of nanoseale technologies, we
consider nanopatteming and nanoasscmbly techniques that employ proteins and D A.
Side R Croup
H\ i /0
Amino G " 7 - - - c\rbOXYIiC Acid Group
H H OH
Figure 5.31 A generic amino acid.
as well as the use of biochemical molecular-recognition principles for the assembly of

nanoscale inorganjc building blocks into macroscopic functional materials.
First, let us consider some of the basic properties of proteins. Proteins are formed by
binding together sequences of a",i"o acids to form chain-like molecules whose "links"
are amino acids. Proteins consisting of only a few amino acids are known as peptides.
Important subgroups of proteins are e"zymes and (ll1/ibodies. Proteins constitute one
of the most important classes of biomolecules and they are sometimes referred to as
the "machines" of the body in view of the large number of diverse functions that they
perform. As mentioned previously, amino acids are the building blocks ofall proteins.
Amino acids are among the most important molecules found in biological systems. In
fact, by binding amino acids together to form chains of amino acids, it is possible to
synthesize all known peptides (short chains of amino acids), all known proteins (long
chains of amino acids), and all antibodies (selected proteins). Antibodies arc proteins
that have high affinities for binding to specific molecules known as allIigens. Thus, an
antibody in a cellular environment has a high probability of binding to its antigcn if
the antibody and antigen are in close proximity. Peptides, proteins, and antibodies are
enormously important in biology. As just one example, proteins function as gates that
control the flow of ions into and out of neurons.
Figure 5.31 illustrates a generic amino acid. All amino acids are derived from the
gcneric structure shown in Fig. 5.31 by V'drying the side group, R. When two amino
acids are in close proximity to each other, the carboxyl group on the end of one amino
acid has an affinity 10 bind to the amino group on the end ofthc other amino acid. When
this bonding occurs, the amino group gives up an H atom, and the carboxyl group gives
lip an OH moiety. The liberated OH and 1-1 form H20 and an amide bond, CO-N}-1.
remains as the bond linking two amino acids. In Fig. 5.32, three of the twenty common
amino acids are depicted. Glycine (G) is the amino acid with the simple side group
of a single hydrogen (H) atom. Arginine (R) and aspartic acid (D) are also shown in
Fig. 5.32. As discussed previously, when amino acids like glycine (G), arginine (R), and
aspartic acid (0) are in water, as is generally the case in biological systems, the amino
(NH z) group on the terminus of one amino acid has an affinity (tendency) for binding
to tllC carboxyl (COOH) group on another amino acid; after such a binding event, the
two amino acids arc bound together by an amide bond (CO-NH bond), and a water
molecule, H2 0, is produced that becomes part of the surrowlding water environment.
NH
II
C NH,
I
NH 0
I II
cn,
on
I
cn
I'
I
c.,,
N1·1 ,
CH,
CH
I
COOH NH 1
1
CH COOl! NH,
ICH COOH
Arginine Glycine Aspartic Acid
(Arg or R) (Gly or G) (Asp or D)
Agure 5.32 Examples or some amino acids.
Such a process provides a very simple example of how chemistry may be used to self-
assemble molecules and structures. By binding a few amino acids together a peptide is
formed. From this discussion it is clear that amino acids are indeed !.he building blocks
of many biomolecules, including peptides, proteins, and antibodies. As explained and
illust'rJ,ted previously, these biol1lolecuJes - peptides, proteins, and antibodies - may be
formed through binding of amino and carboxylic groups at the two ends of the amino
acids. The full set of twenty common amino acids has the following members: alanine
(A), arginine (R), asparagine (N), aspartic acid (D), cysteine (C), glutamine (Q), glycine
(G), glutamic acid (E). histidine (H), isoleucine (I), leucine (L), lysine (K), methionine
(M), phenylalanine (F), proline (P), serine (5), threonine (T), tryptophan (W), tyrosine
(Y), and valine (V).
Just as the binding of NH 2 and COOH groups plays a role in linking many
biomolecules, they may be used to bind a COOI-l-functionalized (COOH-coated) quan-
tum dot to the amino terminus of a peptide, protein, or antibody. In this way, we mimic
a self-assembly technique found in nature as a key step in our arsenal of nanofahrica-
tion techniqucs. II is also possible to use another chemical bond found in nature - the
rhiol bond - to provide a frequently used technique for assembling nanodevicc struc-
tures. Specifically, the amino acid cysteine, depictcd in Fig. 5.33, and scveral other
amino acids (Ievcine, valine, mcthioninc, and serine), have special uses in chemical self-
assembly since each of these amino acids contains a sulfur (5) atom in its side group, R.
Specifically, the sulfur atom has an affinity for binding to a (III) surface of gold and it
binds to quanllun dots like CdS through an S-S or lh;ol bond. These thiol bonds are also
instrumental in the phenomenon of protein folding, where they cause cysteine molecules
at various points along a given protein to be attracted to each other, thus causing protein
folding. Indeed, these S-S bonds are Ihe strongest bonds found of all possible bonds
among amino acids in protein systems.
CH, CII
'\. / '
T cr'
cr' cr' CH,
NH 2 CH
Leveine
COOH NH, CH
Valine
COOH
cr'
S
(Lev or L) (Valor V)
I
sr or cr'
NH,
cr'
CH COOH NH,
cr'
CH COOH NH,
cr'
CH COOH
C)'steinc Serine Methionine
(Cy, or C) (Ser or S) (Met or M)
Figure 5.33 Additional examples of amino acids.
The affinity of sulfur to bind gold is sufficient to result in the wide use of S-Au
bonds to bind molecules to Au surfaces. In the field of molecular electronics, there are
efforts to use molecules as wires, and it is necessary to bind these wires to metal contacts.
By chemically binding a sulfur atom to the end of such a molecular wire, it is possible
to bind the S-functionalized ends of the wire to Au. The use of these 5-Au bonds in
molecular electronics is so prevalent that they are referred to as the alligator clips of
molecular electronics.
Now that we have considered basic properties of proteins, we turn to their applications
in nanofabrication. It is known that proteins form the external surface of many bacteria.
Analogously to the case of block copolymer films analyzed previously, surface~layer
proteins can be used as nanoscale biological masks. As shown schematically in Fig. 5.34,
surface-layer proteins can be isolated from bacterial cells, and subsequently reassembled
on the surfaces of a solid. The surface-layer proteins can be deposited upon various
metals and semiconductor crystals. On their surfaces, surface-layer proteins form two-
dimensional periodic structures of various symmetries. The sizes of the elementary cells
of these surface crystals range from 3 nm to 30 nm. The thickness of an artificially built
surface layer is about 2-8 nm, and it has pores of various sizes (2-8 nm) and shapes. In the
upper part of Fig. 5.35, one such periodic structure is shown. The figure illustrates typical
processing steps for nanopatterning: (a) deposition of surface-layer protein crystals onto
the substrate; (b) metallization by deposition of metal; (c) dry etching, which allows the
transfer of the pattern 10 the substrate; and (d) a plane view of the resulting nanostructure
with highly ordered holes.
Figure 5.34 A schematic drawing of the isolation of surface-layer proteins from bacterial cells
and their reassembly inlo crystalline arrays in suspension at a solid support.
(a)
(b) --:::::::. Metal ovcrlayer

~. Protein
C:::=:=:=:=J
(e)
Hole si esl Dry etching Metal (oxide)
Protein
(d)
. . ~
Hole sites
•••••• Metal (oxide)

on lOp of
protem
Figure 5.35 Processing steps for nanopattcrn transfer: (a) deposition of S-Iayer protein crystals
onlO the substrate; (b) shadow metallization by electron-beam deposition; (e) dry etching to
transfer the pattern to the substrate; and (d) a plane view oCthe idealized nanostructurc. After
T. A. Winningham, S. G. Whipple. et. al., "Pattern transfer from a biological nanomask 10 a
substrate via an intcnnediale transfer layer," J. file. Sci. Technol., 19, 1796--1802 (200 I).
Reprinted with permission from Thomas A. Winningham, Steven G. Whipple, and Kenneth
Douglas, Journal of VaCUli/II Science & Technology B, 19, 1796 (2001).10 2001 AVS The
Science & Technolo!,'Y Society.
Agure 5.36 A scanning clectron microscopy image of GaAs semiconductor wires obtained by
means of the surface·layer protein technolob'Y. The wires have a length of 80 nm and a diameter
of about 811m. Reprinted with permission. from M. Haupt, S. Miller et. al.," anoporous gold
films created using templates formed from self-asscmbled structures or inorganic-block
copolymer micclles," Advanced Mllteriais. 15. 829-831 (2003). Fig. 3(a), p. 83 I.
Then, such a surface "supcrlattice" of holes can be used, for example, to fabricate
semiconductor nanoscale free-standing wires. Figure 5.36 depicts a scanning electron
microscopy image of semiconducting GaAs wires fabricated using surface· layer protein
technology. The wires have a length of 80 nm and a diameter of about 8 nm at the top
of each wire, while the base is thicker. High ordering of free.standing nanoscale wires
is obtained using this biology·based method of nanopatterning.
Biologiealmcthods facilitate the realization of the assembly of nanoscale inorganic
building blocks into functional materials. As an ex.ample, the biological macromolecule
DNA may be used as a synthetically programmable assembler of nanoscale structures
in order to fabricate a wide variety of different ensembles of selected nanocomponents.
Such ensembles may be assembled in a variety of architectures. This method is based on
the molecular-recognition properties associated with DNA. D A may be prepared and
functionalized with virtually any chromophore, acceptor, donor, or an active group in
an automated fashion. It is known that the molecular-recognition properties arise due to
base pairing in DNA strands. Specifically, the commonly encountered double-stranded
D A is composed of two single strands of DNA. These single strands are composed
of linear sequences of the four DNA bases adenine (A), cytosine (C), guanine (G), and
thymine (T). The two single strands bind together to form a single strand as a result of
the high affinities for G and T to bind together and C and A to bind together. Thus, as
a simple example, GTCAC and TGACA bind together to form a double-stranded DNA
molecule. The two single strands, GTCAC and CAGTG. are complementaJy to each
other. These molecular-recognition processes can guide the assembly of nanoparticles
into extended structures. In principle, the method selects nanoparticles with certain
chemical compositions and sizes, and it controls also the distance and coupling between
the particles in the resulting nanostructured materials.
• Au nanoparticJcs
,,~~/ "","'*
Modification with .......~ ~...~ Modification with
3' lhiol TACCGTTG 5' I I 5' AGTCGTTT 3' Lhiol
~!1 Addition orJinking DNA duplex

5' ATGGCAACIIlDTCAGCAAA 5'
B S • -
-
111 ,
<lJ 6
.B
.!it .(11 6 •(8;A•
Further oligomerization
~1I and settling
-'8 - B·
Agure 5.37 Formation of a biomolecule-inorganic c1tl'iter hybrid. Reprinted with permission.

from C. A. Mirkin, "Progrumming the assembly of two- and three-dimensional architectures
with DNA and nanoscale inorganic building blocks:' '"org. Chem., 39, 2258-2272 (2000).
Q American Chemical Society.
As llanoparticle building blocks, nanocrystals of metals, semiconductors, mag-

netic particles, fullerenes, etc. are currently available. The sizes of these nanocrys-
talslnanoparticles can be controlled precisely, in many cases from I nm to 20 nm in
diameter.
The formation of a biomoleclile-il1orgaJlic cluster hybrid is one of the important bio-
logical methods for creati ng nanoscale architectures. This process can be illustrated by the
following example. Let two different sequences of nOli-complementary eight-base-pair
DNA be synthesized with alkane/hiol endgroups. Then, two batches of Au nanoparli-
cles are functionalized with these DNA strands. In the scheme presented in Fig. 5.3 7 this
functionalizing process is shown in the upper part ofthe figure. Two particular sequences
of DNA molecules are indicated as well. If the two batches of Au particles were simply
mixed, there would be no DNA recognition and hence no particle aggregation. However,
the addition of linker DNA strands changes Ihe situation. Such a linker molecule is com-
posed of a DNA molecule Ihat has tluee regions: a central region of double-stranded
DNA and Iwo end regions Ihat are composed of unpaired single strands of DNA. These
unpaired single-stranded ends arc known as sticky e"ds. If the linker DNA molecules
contain eight-base-pair sticky ends that are complemental)! to the base pairs anached
--33nm
Rgure 5.38 A transmission electron microscopy image of a two-dimensional structure of Au
nanoparlicles linked by DNA molecules. Reprinted with permission, from C. A. Mirkin,
"Programming the assembly oflwo- and three-dimensional architectures with DNA and
nanoscalc inorganic building biocks,"/llorg. Chem., 39, 2258-2272 (2000). iOAmerican
Chemical Society.
to the Au particles, the particles starl to aggregate, as shown in the lower part of the
scheme in Fig. 5.37. The process of linking these nanopartic1es with DNA molecules is
known as "oligomerization" since DNA molecules are also known as oligonucleotides.
In this examplc, the so-called sticky ends,S'ATGGCAAC and TCAGCAAA5' arc sin-
glc strands of DNA bound to opposite ends of a segment of double-stranded DNA.
The double·stranded portion of this molecule is represented by the ladder symbol in
Fig. 5.37. As is illustrated in Fig. 5.37, the 5'ATGGCAAC sticky end binds to the
3'thiolTACCGlTG complementary molecule whieh is bound to an Au quantum dOl by
this bond. Likewise, the TCAGCAAA5' sticky end binds 10 the 5'AGTCGTIT3'thiol
complementary molecule which is bound to an Au quantum dot by the thiol bond.
Thus, an object can be built frOI11 nanosized building blocks linked by DNA molecules.
By such a method, both two-dimensional and three-dimensional ordered nanosys-
terns have been fabricated. Figurc 5.38 depicts a transmission electron microscopy
image of a two-dimensional structure of Au nanoparticlcs linked by DNA molecules.
Similarly, semiconductor nanoc1uster-DNA hybrids have been used 10 construct
extended macroscopic structures of CdSeJZnS quantum dots interconnected with DNA
molecules. The wavelength of light emitted by a quantum dot is determined by its size as
well as by the material. Table 5.1 summarizes the bandgaps ofsome direct-bandgap mate-
rials thaI are used to fabricate quantum dots. The materials represented in Table 5.1 have
wavelengths spanning the ultraviolet, visible, and infrared regions ofthe electromagnetic
spectrum.
When we integrate DNA with inorganic building blocks, we can exploit some of the
properties ohhe latter. For example, semiconductor nanocrystals hybridized with DNA
can be used as fluorescent biological labels, because of specific optical properties of
Table 5.1 The bandgap, o!the direct-bandgap

bulk materials used for fabrication of Quantum dots
Compound semiconductor Bandgap (cV)
A1N 6.15
CdS hcxagonal 2.4
CdS cubic 2.55
CdSe hcxagonal 1.75
CdSc cubic 2.17
CdTe 1.49
PbS 0.41
PhSe 0.27
ZnS 3.68
G,N 3.36
these nanosclusters such as their emission at well-defined wavelengths. These optical

properties will be discussed in the next chapters.
Although D A arguably is the most tailorable and versatile molecule for organizing
nalloscale materials into extended structures, its use has some limitations. Most notably,
it is not a high-temperature material, and therefore the structures initially generated from
DNA interconnects will not be stable at elevated temperatures.
As illustrated previously. NH z and CaOH groups playa role in linking many
biomolecules, and may be exploited further in interconnecting ensembles of quantum
dots. Consider the case in which ZnS-coated CdSe quannlm dots (QDs) coated with
carboxyl groups are functionalizcd with GGGC peptidcs. These CdSe-ZnS QDs func-
tionalizcd with GGGC peptides will be denoted simply as CdSe-ZnS-GGGc. This
commonly used notation does not specify the number of peptides bound to each quan-
tum dol. In practical applications, the number of peptides may vary from one to twenty
or more depending upon the application. In the present case, where the outermost amino
acid is a cysteine molecule with its sulfur-containing side group, the C amino acids
may be used for chemical self-assembly. lndeed, if, as the first step in the fabrication
process, an Au wire is immersed in a beaker containing a typical density of 10 16 CdSe-
ZnS-GGGC complexes per cm 3 , many of the CdSe-ZnS-GGGC complexes will bind to
(III) surfaces of the Au wire since the S atoms in the cysteine side groups have affinities
for binding to (Ill) surfaces of Au.ln the next fabrication step, the CdSe-ZnS-GGGC-
coated Au wire is immersed in a beaker containing 10 16 CdS nanocrystals per cm 3 • and
there is subsequent binding of CdS nanocrystals to the C amino acids of the GGGC
biolllolecuies thai are attached to the CdSe-ZnS nanocryst'als Ihat are bound to the Au
wire. Upon repeated alternating immersions of tile quantum-dol functionalized wire in
the two beakers ofCdSe-ZnS-GGGC and CdS, the nanocrystals arc assembled as illus-
trated in Fig. 5.39. The nanocrystals assembled in this manner have densities in excess
of I0 17 per cm]! These densities arc orders of magnitude higher than those achievable by
current commercially used scmiconductor-device-fabrication techniques. This example
illustrates the use of chemical self-assembly to integrate ensembles of semiconductor
nanostructurcs. To create electrically flUlctional integrated semiconductor nanocrystal
••••
I ····
,••••••
••••, CdS
Au wire .....
I·····
•••••
•••••
'!"
-
•••••
';::l ;m;
•••
CdSe-ZnS
I·····
::::1
••••
......
".
~If~. CdSe-ZnS-GGGC
••••••
••••••
••••••
••••
....:I:~
tfl"-
Figure 5.39 A quantwn-dol network assembled using chemically di.rected assembly techniques.
nChl/orks using chemically directed assembly with molecular links, it is necessary to usc
molecular links that conduct electricity.
Since proteins (and short proteins - peptides) constitute onc of the major classes of
biomolecules, an understanding of the electronic properties of these general classes of
molecules is of potential imporlance in bioelectronics. Charge transport in peptides has
recently been studied by inserting amino acids containing a side group containing a
natural chromophore. which is a molecule that produces charge carriers when exposed
to light. In this way, charge can be introduced into the peptide by illuminating the chro-
mophore in thc side group. In addition, phenylalanine, tryptophan, and tyrosine have
been considered theoretically as possible conducting clements of peptides since their
side groups aIC rich in conductive rr-bonds. As wc discusscd for conducting polymers,
such rr-bonds provide charge conduction. Results frolll the chromophorc-based stud-
ies of charge transport indicate that charges move along these peptide-based wires at
speeds of about 1.5 x 105 cm S-l, a relatively slow speed frolll the standpoint of Si-
and GaAs-based electronic devices. This relatively slow transport implies that electronic
systems of integrated nanocrystals with peptide-based interconnects must be designed
with architectures that overcome the limitations imposed by carrier-transport speeds in
peptidcs; perhaps, the design of massively parallel architectures offers a possible reso-
lution of the difficulties imposed by the limited propagation speed in the peptide-based
interconnects.
Our previous discussion of the use of DNA as a self-assembling agent makes it clear
that DNA may be used to link metallic and semiconducting nanostructures. There are also
current research efforts aimed at using DNA as molecular wires that conduct electric
current. Of course. these efforts stem from the growing interest in the use of DNA
as a chargc-carrying element in bioelectronic devices. In these efforts. a key paramcter
determining the change oftransport properties of DNA is the highest occupied molecular
orbital (HOMO). As shown in Fig. 5.40, the commonly encountered type of DNA is a
double-stranded helix-shaped molecule with each of the two strands being composed
of sequences of the following four bases: adenine (A), cytosine (C), guanine (G), and
thymine (T). As discussed previously, these sequences may contain the A, C, G, and T
bases in any order, and the adjacent bases on the two strands are pairs of cither G-C or
Light
E , - - - _ ---::--,----------E IIOMO
h+
Figure 5.40 Transition of a hole from a quantum dOl fO DNA.
A-T. For a simple double strand of DNA that contains only G-e pairs, Ihe energy levels
of the DNA exhibit a gap of the approximate magnitude of2.0eV between the HOMO
and the lowest unoccupied molecular orbital (LUMO).
In molecules, energy slates may be filled, partially filled, or empty just as in semicon-
ductors. As discussed earlier in this book. the valence band of an undopcd semiconductor
is nearly full of electrons at low temperatures, and the conduction band is nearly empty
under these conditions. For molecules, it is traditional to refer to the HOMO simply as
the HOMO "level" or the HOMO band. The highest energy in the HOMO band, there-
fore. plays a role analogous to the high-energy edge, or top. of the valence band £v, of
a semiconductor. Similarly, there is an energy gap between the top of the HOMO band
and the low-energy edge, or bonom, of the next (empty) molecular orbital, known as the
LUMO, which is referred 10 frequently as the LUMO band. The bollom of the LUMO
band plays a role analogous to the low-energy edge of the conduction band, Et .
One of the current approaches to studying charge transport in DNA is to bind DNA
to a semiconductor quantum dot and to inject charge into the quantum dot as shown in
Fig. 5.40. As shown in the lower portion of Fig. 5.40, it is desirable in such experiments
to pick a semiconducting material for the quantum dot that has some of its energy states
aligned with some of the orbitals of the DNA. In the case illustrated in Fig. 5.40, the
valence-band energy is selected so thai it falls just below the HOMO of the DNA. For
a DNA molecule that is composed of a pure strand of G bases bound to a pure strand
ofC bases. the top of the HOMO band has an energy of -7.34 eV as measured relative
to the vacuum level. Two readily available colloidal semiconductor quantum dots with
valence bands having energies close to the HOMO band energy are TiOz and ZnO.
TiOz has conduction- and valence-band energies of -4.21 eV and -7.41 eV relative
to the vacuum level, respectively. ZnO has conduction- and valence-band energies of
-4.19 eV and -7.39 eV relative to the vacuum level, respectively. For either of these
semiconductors, the energy alignment of the valence-band edge with the HOMO band
is as shown in Fig. 5.40. With such a band alignment, holes in the semiconductor may
be injected into the DNA.
In pr<lctical experiments on charge transport using these QD-DNA structures, light

with energy greater than the bandgap of ZnO (3.2 eV) or TiOz (also 3.2 cV) is used to
create electron-hole pairs in the quantum dot. Since ZnO and TiO z are both indirect-
bandgap materials, these electron-hole pairs do not recombine as rapidly as they do in
direct-bandgap materials. (This is one ofthe reasons why TiOz is used in solar cells, where
there is a need to maximize the production of electric currents from photo-produced
carriers.) As shown in Fig. 5.40, the photo-produced holes may escape the quantum dot
and move into the DNA wire. Alternatively, we may view this process as an electron near
the top of the HOMO band making a transition into the quantum dot, where it recombines
with a hole in the valence band. Experiments based upon these techniques are currently
being conducted by several research groups to assess the current-carrying capability
of DNA. One of the interesting findings emerging from such studies is that charges
(holes) tend to become trapped near guanine-rich regions of DNA molecules. This is
not surprising, since the ionization potentials ofGC and AT base pairs are -7.34 eV
and -7.99 ev' respectively. Since these ionjzation potentials correspond to the cnergies
required for an electron to transition from the top of the HOMO band to the (unbound)
continuum, it follows that the HOMO band·edge energy has a local maximum near
guanine-rich sites along the DNA. Accordingly, a hole that is propagating along the
D A wire and dissipating its energy as it propagates - by producing vibrational modes
in the DNA, as an example - may become trapped at the guanine-rich sites. The charge-
transporl properties of DNA are still not understood fully. However, if conducting DNA
wires may be engineered by selecting appropriate base sequences, it may be possible to
use DNA not just as a self-assembling agent of complex ensembles of nanocomponents.
but also as an electrically active linking clements in these ensembles ofnanostructures.lt
may even be possible to design novel types of ultra-high-density information-processing
systems that greatly exceed the maximum infonnation-processing capabilities realizable
by the downscaling oftoday"s integrated circuits! At present, it is reasonable to expect
that years of research will be required to determine whether such nanostructure-based
chemically self-assembled systems are useful for advanced information processing.
Dip-pen nanolilhography
Since the invention of the scanning tunneling microscope and similar technjques, there
have been attempts to develop new nanolithography methods. In particular, STM- and
AFM·based methods were applied to oxidize, scrape, or etch nanostmctures on surfaces.
Some ofthese methods were mentioned in previous sections. However, proposed methods
are generally limited to the growth of thin oxides on selected metal and semiconductor
surfaces, or to multistep etch procedures that cannot be generalized to parallel and
productive patterning of nanostruclures.
Finally, dip-pen nanolithography (DPN) was introduced as a direct-write scanning-
probe-based lithography in which an AFM tip is used to deliver chemical reagents directly
to Ilanoscopic regions of a target substrate, as shown in Fig. 5.41.
The DPN technique is a type of soft lithography, where the word "soft" refers to the
chemical composition of the nanostructures which can be fabricated. They are made
J • Molecule to be
Tip with Nanoscale r---- Assembled
Dimensions •
Molecular
Transport
Writing Direction
!!!!!!!I 'I 7 ~
~~~
Figure 5.41 The principal "writing" clcment of a dip-pcn lithographic sys(cm is a nanoscale
structure having a tip with nanoscalc dimcnsions. Fluids flowing ovcr thc surfacc ofthc tip are
dcposited onto the surface of a substratc in the region where a meniscus is formcd between the
tip and the surface. Reprinted with permission. from C. A. Mirkin. "Progrnmming the assembly
of two· and three-dimensional architecturcs with D A and nanoscale inorganic building
blocks:' Illorg. Chem.. 39. 2258-2272 (2000). <0 American Chemical Society.
of organic ligands rather than solid-state materials. Though it is unlikely that dip-pen
technologies can displace conventional solid-state fabrication methods, they are highly
complementary. Many interesting scientific and important practical issues pertaining to
miniaturization can be addressed as a result of having these soft lithography methods,
whereby molecules can be patterned in a controlled fashion on the sub-I OO·nm length
scale. For example, such a method can generate nanoscale 1110lecule·based conduct-
ing structures and provide their contacting with macroscopically addressable electrodes
prepared via conventionalmicrofabrication methods.
In inventing DPN, the idea was to miniaturize a 4000·year-old tcchnology: the technol-
ogy ofthc quill or dip-pen. The largest difference is that we wanted to do on a nanoseopic
scale what a quill pen can do on a macroscopic scale.
Considering the fundamentals of DPN, first we should mention that the simple idea
of transporting any "ink" through a nanoscale AFM tip to a surface via the well-known
capillary action does not work properly. instead, the basic idea is to design a system with
an ink that would chemically reacI with a substrate of interest. The scheme in Fig. 5.41
represents the main DPN clements: the moving AFM tip is coated with ink molecules;
a water menisclls forms between the tip and the solid substrate; ink molecules react
with the substrate. Such an effect of reaction of a chemicaJ with a surf.,cc is called
chemisO/pliol/. Thus chemisorption acts as a driving force for moving thc molecules
from the tip to the substrate and then results in formation of stable one-molecule-thick
nanostructures. The key to making this approach successful is to select organic molecules
with low water solubilities so that the chemisorption driving force facilitates the control
of tip.to-substrate transport properties. This would prevent the uncontrolled, nonspecific
adsorption and accumulation of multi layers of molecules on the surface.
Figure 5.42 illustrates the results of chemisorption. The panel on the left shows the
coating - "painting" - ofa 1_~m2 area of the gold surface with octadecanethiol via the
use of DPN. The formation of a one-monolayer structure is proved by the more detailed
5.9 Fabrication of nanoelectromechanical systems 157
(a) (b)
o 11m 3 o A 150
Figure 5.42 (a) The paiming via DPN by octadecanethiol on gold. (b) A Janice-resolved image of
all octadecanethiol nanoslructure deposited on a single crystal of gold. Reprinted with
permission, from C. A. Mirkin. "Programming the assembly of 1\\10- and three-dimensional
architectures with D A and nanoscalc inorganic building blocks,"/norg. Chem., 39, 2258-2272
(2000). American Chemical Society.
lattice-resolved image of the transported octadecanethiol, as depicted in the panel on the

right of Fig. 5.42. The observed lanice is hexagonal with an intermolecular spacing of
5 A, the known lattice constant for an octadecancthiol monolayer on gold formed by a
number of other methods.
The application of the method is shown in Fig. 5.43. The panel on the left shows a
molecular grid of octadecanethiol on gold. The grid is a single molecule thick. with 100-
nm-wide lines. The panel on the right demonstrates an array of uniformly deposited 450-
nm-diameter dots. Note that attaining this type of shape regularity and reproducibility
would not be possible with a conventional pen, even over macroscopic dimensions.
Currently, writing with DPN generates a nanostructure with approximately 60-70 nm
linewidths. The smallest fabricated structurcs are 15-nm-diametcr dots spaced
5 nm apart. DPN has been developed to pattern a variety of ink-substrate combinations.
The method is compatible with many inks. from small organjc molecules to organic
and biological polymers and from colloidal particles to metal ions and sols; patterned
surfaces range from metals to insulators and semiconductors. In Table 5.2. examples of
combinations of chemical inks and surfaces are listed.
In conclusion. chemistry and biological methods can be applied for fabrication of
nanosystems. Importantly, these new methods can be complementary with the conven-
tional semiconductor nunotechnologies.
5.9 Fabrication of nanoelectromechanical systems
In previous sections, we studied state-of-the-art semiconductor technologies Ihat may be

used to produce nanoscale structures and devices for electronics. High-quality structures
Table 5.2 Examples of ink-substrate combinations used in OPN
Ink Substrale Noles
Alkylthiols Au J 5 nm resolUlion on single-crystalline surfaces,

<50 nm on polycrystallinc surfaces
Ferroccnylthiols Au Redox-active nanoslructures
Silazancs SiO. Patterning 011 oxides
GaAs
Proteins Au, SiO.. Both direct write and indireci assembly
Conjugated polymers SiO.. Polymer deposition verified spectroscopically
and electrochemically
DNA Au, SiO.. Sensilive to humidity and tip-silanization conditions
Fluoresccnt dyes SiO.. Luminescence patterns
Metal salts Si, Ge Electrochemical and eleclrolytie deposition
Colloidal particles SiO.. Viscous solulion pallemed from tip
(b)
o 4 o 5
Figure 5.43 The nanoscale grid (len) and Ihe dol army (right) generated via DPN. Reprinted with
permission, from C. A. Mirkin. "Programming the assembly of two- and three-dimensional
architectures with DNA and nanoscale inorganic building blocks:' II/org. Chem., 39, 2258-2272
(2000). ClAmerican Chemical Sociely.
were necessary in order to obtain superior electronic properties. The great achievements
and the advances brought abollt in mainstream electronics by these technologies can be
used for the exploration of other nanodevices commonly called nanoeleclromechanical
sysfems (NEMSs). This class of devices includes nanomachincs, novel sensors, and a
variety of new devices that function on the nanoscale.
Nanomechanical devices promise to revolutionize measurements of extremcly small
displacements and extremely weak forces, particularly at the molecular scale. Indeed
with surface and bulk nanomachining techniques, NEMSs can now be built with masses
approaching a few attograms (I attogram = 10- 18 g) and with cross-sections of about
5.9 Fabrication of nanoelectromechanical systems 159
Electrical Electrical
-
inpul2ignal output signal
• •-
~ M~
Input Output
transducer transducer
Mechamcal
stimulus response
MeChanicaI
• _ perturbation
Electrical
control signal
II" Control
transducer
Figure 5.44 A schematic representation of a three-terminal electromechanical device.
10 nm. The small mass and size of NEMSs give them a number of unique attributes that
offer immense potential for new applications and fundamental measurements. In general,
the potential applications ofNEMSs are likely 10 be enormous and could benefit a diverse
range of fields, ranging from nanoelcctronics to medicine and biotechnology. In this
section, we study the basic concepts ofNEMSs, the technology for NEMS fabrication.
and the challenges arising in this field.
An electromechanical device can be thought of as a two·, three-, or, generally, multi·
terminal transducer that provides input stimuli (i.e., signal forces), and reads out a
mechanical response (i.e., output displacement). At additional control terminals, elec-
tric signals can be applied and subsequently converted by the control transducers into
varying forces to perturb the properties of the mechanical element in a controlled
and useful manner. The generic picture of a NEMS is shown in Fig. 5.44, where
the input, output, and control transducers/terminals are presented schcmatically. The
basic mechanical element of a NEMS is a nanosize suspended film. a membrane, or a
beam; in the following discussion, we will use the term "beam." Easy flexural deforma-
liolls of such mechanical elements provide high mechanical responsivity for a NEMS.
Electronic devices to which the beam is coupled are assumed to be of comparable
dimensions.
Let us consider the fabrication of the basic mechanical component of a NEMS. The
crystal and hetcrostrucnlre growth, and processing tcchniques, which have been stud-
ied previously are used widely to produce suspended semiconductor structures. These
techniques can be applied to bulk silicon, epitaxial silicon. and systems based on III-V
compounds and to other materials.
The procedure for fabricating a suspended structure is illustrated in Fig. 5.45. In
its simplest form, the procedure starts with a heterostructure that contains structural
(a) (b)
(e) (d)
Agure 5.45 A general scheme for the fabrication of NEMS suspended structures using structural
(gray) and sacrificial (dark) layers on a substrate (dark gray). (a) Three-layer base
heterostructure. (b) etching-mask deposition, (e) anisotropic etching. and (d) selective wei
etching or the sacrificial layer.
(gray) and sacrificial (dark) layers on a substrate (dark gray), as in Fig. 5.45(a). Masks
on the top of this heterostructure can be patterned by a combination of optical and
electron-beam lithography, followed by thin-film-deposition processes. The resulting
mask protects the material from beneath during the next stage; see Fig. 5.4S(b). Unpro-
tected material around the mask is then etched away using a plasma etching process
as in Fig. 5.45(c). Finally, a local chemically selective etching step removes the sac-
rificial layer from the specific regions to create a freely suspended nanostruclure Ihat
is both thermally and mechanically isolatcd; see Fig. 5.45(d). This proccdure can be
repeated several timcs and combined with various dcposition processcs to produce
the necessary mechanical nanostrllcture for a particular device. The flexibility of the
process allows one to apply this general scheme to diverse materials and to fabri-
cate fully suspended structures with lateral dimensions of approximately a few tens of
nanometers.
Consider, for example, the important case of silicon nanomachining by lhe use of
the so-called SIMOX (Separation by IMplantation of OXygen) process. The procedure
starts from a Si wafer, which is processed by a large dose of oxygen-ion implantation.
The implanted wafer is annealed at high temperature to form a SiOz layer of 0.05-1 }-lm.
Above the SiOz layer formed, a single Si crystal layer is then grown. This top layer is 0.1-
0.2 }-lm thick. As the result, one obtains a SilSiOz/Si heterostructure that corresponds
Figure 5.46 Four different NEMSs fabricated by SIMOX method. From A. N. Cleland,
Foundations ofNallomechanics. Fig. 11.4 (Berlin, Springer- Verlag. 2003).
to the gray/dark/dark gray structure of Fig. 5.45. All stages are already illustrated in
Fig. 5.45: the top Si layer is patterned and an etching mask is created as in Fig. 5.45(b);
to the top Si layer anisotropic etching is applied as shown in Fig. 5.45(c); then, the oxide
is subjected to selective wet etching to create finally a suspended structure as depicted
in Fig. 5.45(d). The SIMOX process can be carried out with wafers of large area (4 to
6 inches in djameter) and facilitates the integration of a numbcr of NEMSs and other
electronic devices on a chip. Figure 5.46 illustrates somc suspended structures fabricated
by this method.
In conclusion, due to advanced technologies, fabrication of new nanoelectromcchani-
cal systems becomes possible. These systems have dimensions so small that theirmechan-
ical motion (vibrations) are coupled to the electrons much more strongly than in the case
of bulk-like, massive samples. A number of nanodcvices can be built on the basis of
NEMSs. Chapter 8 provides additional discussion of these devices.
5.10 Closing remarks
In this chapter, we focussed on methods used for high-quality material growth and for
nanodevice fabrication. We studied the growth of perfect crystals and multilayered het-
crostructures. We found that there has been a considerable and persistent improvement
of traditional semiconductor technologies for material processing. We also found that
these advances facilitate the fabrication of nanostructures and nanodevices with excel-
lent precision and high reproducibility, and with the necessary electrical, optical, and
mechanical properties.
Moreover, we analyzed novel approaches to producing nanostructures based on special
regimes ofmaterial growth (the Stranski-Krastanow regime), in which nanostructurcs arc
formed spontaneously due to the growth kinetics. The processes of self-organization and
sclf.ordering ofnanostructures (quantum dots. quantum wires, etc.) give rise to a new and
unique way to produce high-density ensembles of nanodcvices with desired properties.
Apart from the traditional methods of fabrication. those adapted for microelectronics
and nanoelectronics, we studied fabrication techniques for such "original" nano-objects

as nanotubes.
We described both improved and novel characterization and growth methods thai
facilitate the control ofnanostructure geometries with atomic-scale precision as well as
the manipulation ofa single atom or ion in a crystal.
The latest achievements in Ihe chemical and biological sciences have generated new
approaches to nanofabrication that arc complementary to conventional semiconduc-
tor technologies. We presented several examples of the ncw chemical and biological
methods.
Finally, we studied in detail the fabrication of a new class of nanodevices, nanoelec-
tromechanical systems. Such nanosized systems are so small that the motion of a few
electrons may strongly affect the mechanical vibrations of nanostructures.
More information on crystal growth and device fabrication can be found in the fol-
lowing books and papers:
R. H. Hendel, S. S. Pei et 01., "Molecular-beam epitaxy and the technology of

selectivel y-doped heterostructure transistors," in Gallium Arsenide Technology, ed.
D. K. Ferry (Indianapolis, IN, Howard W. Sams, 1985).
K. Ploog, "Delta-doping in MBE grown GaAs: concept and device application," 1
OJ'sr. Grow/It, 81, 304 (1987).
D. Bimberg, M. Grundman, and N. N. Ledentsov, Quantllm DOl f-Ieterostl1lctures
(Chichester, John Wiley & Sons, 1999).
TI1C following publications are devoted to sclf-organizcd growth of nanostruehires:
R. Notzel, "Self-organizing growth of quantum dot structures," Sem;cond. Sci. Tech-

1/0/.11,1365 (1996).
V. A. Shchukin and D. Bimberg, "Spontaneous ordering of nanostructures on crystal
surfaces,'· Rev. Mod. Pltys., 71, 1125 (1999).
J. Stangl, V. Hoi, and G. Bauer, "Structural properties of self-organized semiconductor
nanostructures," Rev. Mod. Pllys., 76, 725 (2004).
Scanning tunneling and atomic force microscopy techniques are described in the book
C. J. Chen. 'ntroductioll to Scanning Tunneling Microscopy (New York, Oxford Uni-

versity Press, 1993).
Detailed analysis of fabrication methods of carbon and other nanotubcs is presented in
M. S. Dressellhaus, G. Dressellhaus, and P. C. EkJund, Science of fitl'erenes and

Carbon NmlOtubes (San Diego, CA, Academic Press, 1996).
In the following two publications the chemical and biological nanofabrication methods
are reviewed:
C. A. Mirkin, "Programming the assembly oftwo- and three-dimensional architectures

with DNA and nanoscaJe inorganic building blocks,"'norg. Chem., 39. 2258-2272
(2000).
5.11 Problems 163
D. S. Ginger, H. Zhang, and C. A. Mirkin, "'The evolution ofdip-pen nanolithography;'

Angewa"dte Chem., 43, 30-45 (2004).
A description of fabrication of various types of NEMSs is presented in
A. N. Cleland, Foundations ojNallomecJumics (Berlin, Springer-Verlag, 2003).
5.11 Problems
1. Describe the main differences between the Czochralski method of crystal growth
and epitaxial growth. Which of these approaches can be applied to grow multilayered
crystalline structures?
2. In the molecular-beam epitaxy method, the rate of crystal growth is characterized

by the flux density, J. of atoms, constituting the growing film. For binary crystals AS,
like GaAs, SiGe, elc., the fluxes of deposited components A and B arc to be equal:
J A = Ja = J /2. Let the crystal density PAB be given. Using given J. PAB, and masses of
atoms A and S, calculate the time necessary to grow a filmofthicknessd. Estimate numer-
ically the growth time for GaAs film of thickness 100 nm at J = 10 15 atoms cm-2 S-I;
(PGoA' = 5.316 g cm-').
3. When a photolithographic method is applied.. the effect of diffraction oflight restricts

the minimal scale of an illuminated pattern necessary for further processing to fabricate
a nanostructure. Explain the advantages in the use of short-wavelength light sources in
nanolilhography.
Assume that the minimal thickness of a light line d rnin is related to the wavelength
of illumination A as d min ~ A/2. Calculate and compare the minimal scales of devices
fabricated by exploitation of three laser sources: a red He-Ne laser (A = 0.63 IJ-m), a UV
KrF laser (A = 0.243 IJ-m), and a UV ArF laser (,)" = 0.19 }.un).
4. Apply the "hydrogen model" of Eq. (5.4) for energy levels of donors in GaAs and
InAs. Use the effective masses according to Table 4.5 and sel the dielectric constants, e,
equal 10 12.8 and 15.5, respectively. Calculate the ionization energies and estimate the
radius of ground donor states for these materials. Find the average number of primitive
cells ';covered" by a single donor electron.
5. Explain the role of lattice mismatch for regimes of self-organizing growth of nanos-
tnlctures. For which of the following heterostructures is the formation of nano-islands
•
possible: GaAs/AIAs (the lattice constants arc Go = 5.64 and 5.66 A, respectively),
• •
GaAsilnAs (ao = 5.64 and 6.06 A), and Si/Gc (ao = 5.43 and 5.65 A).
6. Discuss the differences between the working principles of the nanoscale techniques
of scanning tunneling microscopy and atomic-force microscopy.
7. Scanning tunneling microscopy (STM) is a new and important technique
used to probe and characterize nanostructures. Explain why atomic-scale precision
measurements are possible when a macroscopic tip of dimension about I ~m is used.

What are the limitations for applications of STM?
8. Some chemical and biological methods of formation of surface nanopatterns can

be complementary 10 standard semiconductor technologies. Which of these meth-
ods can be used for creation of periodic patterns and which are for arbitrary surface
patterning?
6 Electron transport in semiconductors
and nanostructures
6.1 Introduction
In previous chapters we studied advances in materials growth and nanostructure fabri-

cation. In the case of electrons, we paid primary anention to the quantization of their
energy in nanostructures. In fact, electronics relics upon electric signals, i.e.. it deals with
measurements of the electric current and voltage. Controlling and processing electric
signals are the major functions of electronic devices. Correspondingly, our next task will
be the study of transport of charge carriers, which are responsible for electric currents
through nanostructures.
The possible transport regimes of the electrons are dependent on many parameters and
factors. Some important aspects of these regimes can be elucidated by comparing the
time and length scales of the carriers with device dimensions and device tcmporal phe-
nomena related to operating frequencies. Such an analysis is carried out in Scction 6.2.
In Sections 6.3 and 6.4 we discuss the role of electron statistics in transport effects.
Theil, we consider the behavior of the electrons in high electric field including so·called
hor-elecrron effiers. Analyzing very short devices, we describe dissipative transport and
the velocity·overshoot effect. Finally, we consider semiclassical ballistic motion of the
electrons and present ideas on quantum transport in nanoscale devices in Section 6.5.
6.2 Time and length scales of the electrons in solids
We start with an analysis of possible transport regimes of the electrons in nanostructures.

Since there is a large number of transport regimes, we introduce their classification in
terms of characteristic times and lengths fundamentally inherent to electron motion.
Electron fundamental lengths in solids

As we already noted in Section 4.3, the characteristic length in a crystalline solid is
the lattice constant aQ. However, the scales which are relevant to the charge carriers are
typically much larger than aQ_ This fact, as stressed in Chapter 4, allows one to neglect
fine crystalline structure and consider an electron as an almost free particle, in particular
by assigning to the electron an effective mass that may differ from the mass of an electron
III vacuum.
166 Electron transport
800
T= 300 K
-"~ 400
~
g"
g 200
-L.U
---------------~--
o
10-2 10- 1 100 10 1
Dimensionless electron effective mass (m'/mo)
Rgure 6.1 Electron wavelength versus the electron effective mass for room temperature
(T = 300 K). Points I through 4 correspond to InSb. GaAs. GaN. and SiC, respectively.
The first fundamentally important length is the de Broglie wave/engr" of an electron

in a solid. For a free particle this length was introduced in Chapter 2; see Eq. (2.45). For
an electron in a semiconductor nanostructure with the effective mass m* the de Broglie
wavelength A is typically greater than that of a free electron AO:
A = 2Tdl = 21ftl = Aojm o . AO = 21ft; . (6.1)

p J2m· E m· J2moE
where £ is the electron energy and mo is the mass of the electron in vacuum. In Fig. 6.1
the value Ais shown as a function of",· / "'0. Points 1-4 on the curve indicate wavelengths
for electrons in luSb, GaAs, GaN, and SiC, respectively. We have used effective masses
m* / "'0 equal to 0.014, 0.067, 0.172, and 0.41, respectively, for these materials and we
have assumed that the electron energy is E = ka T. Here T = 300 K is the ambient
temperature and kB is Boltzmann's constant. We see that the de Broglie wavelength of
an electron ,in typical semiconductors with 11/. in the range (0.01-1)11/0 is of the order
of 730-73 A; i.e., it is really much larger than the lattice constants for the materials
presented in Table .... 8. As the temperature decreases to 3 K, the de Broglie wavelength
increases by one order of magnitude. Thus the wavelength becomes comparable to the
sizes of semiconductor structures and devices fabricated by modern nanofabrication
technology.
Size of a device and electron spectrum quantization

Let us introduce a geometrical size of a semiconductor sample Lx x L y x L=, as shown
schematically in Fig. 6.2. Without loss of generality we assume that L: < L), < Lx. If
6.2 TIme and length scales 167
Contacts
, ,
, ,
L.
<al (b)
Figure 6.2 (a) Geometrical sizes of a semiconduClor sample (L= < L) < Lx) and (b) a sample
with contacts; the electron transport occurs along the x-direction.
the system is free of randomness and other scattering mechanisms are sufficiently weak,
the electron motion is qllasiballisric and the only length with which the geometrical sizes
need be compared is the electron de Broglie wavelength A. Since only an integer number
of half-waves ofthe electrons can fit into any finite system, instead ofa continuous energy
spectrum and a continuous numherofthe electron states, one ohtainsa set ofdiscrete elec-
tron states and energy levels, each ofwhich is characterized by the corresponding number
of half-wavelengths. This is frequently referred to as quantization of electro" motioll.
Depending on the dimensions of the system, one can distinguish the following cases.
(a) The three-dimensional or bulk-like case, when the electron spectrum quantization is
not important at all,
(6.2)
and an electron behaves like a free particle characterized by the effective mass m*.
(b) The two-dimensional or quantum-well case, when the quantization of the electron
motion occurs in one direction while in the other two directions electron motion is
free:
(6.3)
Such a case was discussed in Chapter 3 for the example of a potential energy depen-
dent on a single coordinate. The electron energy for this case is given in the form of
two-dimensional subbands, as described by Eq. (3.49).
(c) The one-dimensional or quantum-wirc case, whcn the quantization occurs in two
directions, so that the electron moves freely only in one direction - along the wire:
L, '" L y "'!. «L.,. (6.4)
Such a case was disclissed in Chapter 3 for the example of a potential energy depen-
dent on two coordinates. The electron energy for this case is given in the form of
one-dimensional subbands, as in Eq. (3.50).
(d) The zero-dimensional or quantum-box (quantum-dot) case, when the quantization

occurs in all three directions and the electron can not move freely in any direction:
(6.5)
The simplified models for this case also were analyzed in Chapter 3. The energy
spectrum is discrete.
The last three cases also illustrate the quanfum size effects in one, two, and three
dimensions, respectively. If at least one geometrical size of a device is comparable to the
electron wavelength, a quantum-mechanical trealment of the problem is strictly required.
Let us analyze the conditions when and the reasons why carriers lose their wave-
like behavior so thai they can be considered as classical particles. There arc two major
reasons. The first is non-ideality of the system, which leads to electron scattering. The
second is related to finite temperature and electron statistics.
Electrons in solid-state devices are subjected to scattering by crystal imperfections,
impurities, lattice vibrations, intcrface roughncss, etc. These scattering processes are
divided into rlVo groups: elastic and inelastic. In classical physics, an elastic collision
leads to a change only in the particle momentum (wavevector), whereas in an inelastic
collision both the momentum and the energy change. An essential property of an elastic
collision is that it does not desrroy rhe phase oJ the electron. Indeed. after an elastic
scattering the energy remains unchanged and the electron wavefunction l.JI(r. t) consists
ofdifferent components, which are oftype e- iOI el~r. All components have the same time-
dependent phase e- iO,. Thus, the spatial distribution of the electron density J l.JI(r. t)1 2 =
11ft(r)1 2 remains independent of time. In other words, elastic scattering does not destroy
tile coherence of electron motion. The same is true for the case of two or more impurities:
thc spatial wavc pattcrn is generally complcx, but it remains coherent.
Using the semiclassical language, if T e refers to the mean time between two elastic
scattering events, we can define the mean free parh of the electrons between elastic
scattering events as Ie = UTe, where u is the average electron velocity. l1lerefore, even
for distances exceeding Ie the wave-like properties of electrons are coherent.
Inelastic scattering leads to a new result. This scattering produces electron waves
with different energies and the resulting wavefunction has a complex dependence on
both position and time; the beating of different wave components in time washes out
the coherence effects. Let TE be the mean time between two inelastic collisions. The
distance the electron propagates between these collisions is frequently called the inelasric
sCllllering lenglh. LF... The electron preserves its quantum coherence for distances less
than LE and it loses coherence for larger distances. Generally, LE > A unless extremely
nonequilibrium conditions exist. Often L E far exceeds the mean free path, Ie. In this case,
the electron lmdergoes many elastic collisions before losing its energy. This process is
known as diffusion and its displacement during TE is known to be
(66)
where the diffusion coefficient D is given by D = U 2 Te/ex, with ex = 3 for a three-
dimensional electron gas, ex = 2 for a two-dimensional electron gas, and ex = I for a
6.2 Time and length scales 169
one-dimensional electron gas. Usually, tE and iE decrease as the temperature of the

system increases.
The second reason for averaging out the quantum behavior is the temperature effect on
the electron statistics. Indeed, at finite lemperahlres, there are electrons with significantly
different energies and this leads 10 a large spreading of the wavcfunction phases and the
coherence in the electron system is destroyed. One can estimate the characteristic length
iT associated with such a temperature-related smearing of the coherence. Electron
energy broadening of the order of kB T leads 10 a spreading of the phases with time,
t. Ihat goes as 6¢J = t X (ke T ItJ). TIlUS, the time of spreading. t. can be estimated as
the instant fr, for which 6¢J ~ I. i.e., tr = fll(ksT). If the only scattering is elastic
scattering. an electron diffuses in space over a distance of about JDi during the time t
exceeding the mean-free-flight time te:' Therefore, during the time interval tr one obtains
J
the thermal diffusion length L T = J Dtr = Dff/(ke T). At distances exceeding Lr,
the coherence of the electron will be lost.
In fact. the dephasing effects caused by inelastic collisions and the temperature spread-
ing of phases exist simultaneously. The spatial scale associated with the loss ofquantum-
mechanical coherence should be determined by the smaller of these two lengths:
141 ~ min(i e . ir}· (6.7)
Electron transport is determined by the wave function. i.e.• by the superposition of the
scattered electron waves. From the above consideralions we can conclude that the coher-
ence length. 'IP.
defines the limit below which electron transport has a quantum character.
Devices with geometrical sizes of the order of the coherence length arc no longer char-
acterized by macroscopic material parameters such as conductivity, average velocity,
etc. Such systems are called mesoscopic systems. The proper theory for the description
of mesoscopic devices is therefore quantum theory; the properties of such mesoscopic
systems are determined by wave-like phenomena and are thus strongly dependent on
the geometry of the sample, contacts, positions of scat'lerers, etc. For cases in which the
transport distance, L:, is long compared with I~, the device can be described within the
framework of classical physics.
Quantum and classical regimes of electron transport

Let us compare the previous discussion offundamentallengths with characteristic device
sizes to illustrate and explain possible electron-transport regimes. For simplicity, we
suppose that transport occurs along one dimension, say the x direction. The total current
in each of the other Iwo directions is zero, but these transverse sizes of the device can be
Important too.
Quantum and mesoscopic regimes of transport

We can define two non-classical regimes. If the de Broglie wavelength exceeds the length
ofa device in
(6.8)
and Ie »At electron-transport is described in terms of the {julIll/um ballistic transport

regime. If the coherence length./~ (also referred to frequently as the dephasing length),
exceeds Lx and At
(6.9)
electron transport is described in terms of tile mesoscopic frat/sport regime.
The classical transport regime

In the case when the size L.t exceeds the dephasing length,
(6.10)
electron transport is described in terms of the classical regime. If the dimension Lx is

less than the mean free path,
(6.11 )
electron transport is described in terms of the classical ballistic regime, which means
that electrons can move through the device along classical trajectories without collisions.
If the dimension L.r is greater than the mean free path,
(6.12)
electron transport is of a diffusive nature. If L., --IE» Ie. electrons do not lose their
energy in moving across the device. Such transport is called quasiballistic transport. In
the absence of an electric ficld the electrons preserve their energy under the quasiballistic
regime. On combining the above-discussed inequalities with inequalities (6.2)-(6.4), one
can see that there are three classical trnnsport regimes for one, two, and three·dimcnsional
electrons.
Ifthe transverse dimensions, L; and L y, are both greater than the de Broglie wavelength
but they are comparable to one of the characteristic classical lengths, the transport regime
is characterized by transverse classical size effects. In this case, collisions with the device
boundaries affect electron transport through the device. For example. if one or both of
the transverse dimensions are of the order of the mean free path,
the resistance of the device depends strongly on the properties of the side boundaries.
Roughness of the boundaries increases the resistance and entirely controls it if L;. L y «
I,.
Iflhe transverse dimensions become comparable to one of the diffusion lengths, we are
dealing with another kind of classical size effect, namely, diffusive classical size effects.
For example, if L: or L J' is of the order of the energy relaxation length L I~ the device
boundaries provide an additional energy-relaxation channel. This diffusive size effect
controls the mean energy of nonequilibrium electrons. For convenience the classification
of the possible transport regimes is presented in Table 6.1.
6.2 TIme and length scales 171
Table 6.1 Classification of transport regimes
Intercontact distancc, L", is comparable to

Quantum rcgime the clcctron wavclength. L~ :5 ).
Mesoscopie regime Intercontact distance is less than the dephasing length.

LA < /.
Classical regime (one-, two·, and lntercontact distance exceeds the dephasing length,
three-dimensional electron L" > ,_:
transport) classical ballistic regime, Ie > L~
quasiballistic regime (enerh'Y·conserving):
L E ~ L.• ~ le./.
transverse size effects:
effect related to the mean free path, L=. L f -I.
diffusion effects. L:. L, ....... L E
Time scales and temporal (frequency) regimes

The time scales which characterize transport phenomena determine the temporal and
frequency properties of materials and devices.
There are two fundamental times defining the character of electron-transport behavior:
the time between two successive scattering event's, or the free-flight (scattering) time, Tc;
and the time which characterizes the duration of a scallering event, 1 s. Under ordinary
conditions 1 c » 1$' In fact, it is usually assumed that the scattering event is instanta-
neous, i.e.• 1 s ~ O. In this case either classical or quantum theory can be applied for
the description of electron behavior, depending on the length scales. If. however, 1 c is
comparable to or smaller than 1 s, which may happen WIder extremely strong scattering
of nonequilibrium electrons, the quantum description of electron behavior is required
regardless of the size of the system.
In classical transport regimes, the characteristic times and their relationships to the
device sizes are of critical importance. They determine temponl1 and frequency regimes
of device operation. For example, the ftansit time fIr = L.rlv determines the duration of
signal propagation through a device; here 1I is the electron velocity. Therefore, fir defines
the ultimate speed limit of the device: the device cannot effectively operate in the time
range less than fir or at frequencies greater than t1;1. This explains one of the trends of
modern electronics: scaling down the device sizes.
The times related to transverse dimensions, (b = L: ..I.jlJ (near the ballistic regimes),
or to = (L:. y )2/ D (for diffusive size effects), determine features of electron transport at
frequencies of the order of tbl, or rj)l.
In quantum mechanics, as described in Chapter 3, if external potentials are time-
independent, the electrons are in srafionmysrates. In this case, despite a possible complex
dependence of the wave function on position, the temporal evolution of a stationary state
is always determined by an exponential factor exp[ -iCE /tI)f]. If the alternating external
field of an angular frequency w is applied to the stationary electron system the response
of the electron system may be referred to one of the following three different regimes
depending on the frequency of the external field.
Ultra-high (quantum) frequencies

If tun is comparable to the characteristic stationary electron energy state, E, the nature
of the electron's response will be essentially quantum· mechanical. Only transitions
behvecn states with the energy difference 6.E = 1lw arc allowed. If £ is quantized,
the interaction is possible only at resonance frequencies. By varying the device size,
one can vary the energy spectrwn and, as a result, change the frequency properties over
a wide range. The kinetic times T~. TE. etc. lead 10 a broadening of these resonances.
If this broadening exceeds the energy separations between quantized levels, the dis-
crete quantum behavior changes to a continuum-like behavior reminiscent of classical
mechanics.
Ift,w« £ the electron's response to an alternating field is classical (quantization of
transitions can be neglected). In the classical picture, the external alternating field will
cause periodic electron acceleration and deceleration. The scattering interrupts these
accelerations and decelerations. Depending on how many scattering events occur during
the one period, we distinguish two diffcrent regimes of electron behavior.
High (classical) frequencies

If wT( » I the electron motion during one period is not interrupted by scattering. In
accordance with classical mechanics, the electron momentum oscillates with a phase
opposite to that of the field.
Low frequencies
If WTe « 1 the electron undergoes many scattering events during onc period of the
external field. Multiple scattering during the period brings the electron into a quasi-
stationary state, which follows the oscillations of the external field. In other words, the
electron momentum oscillates in phase with the field.
In closing this section we conclude that depending on the device dimensions, the
temperature, and other conditions there is a variety of transport regimes. Each of these
regimes demonstrates peculiar properties and requires a physical description suited to
the relevant conditions.
6.3 Statistics of the electrons in solids and nanostructures
For further analysis of nanostructures, we need 10 review briefly the basic properties of
many-electron systems. Indeed, any semiconductor material consists of a vast number
of electrons, i.e., it is essentially a l1umy-electroll system. For such a system, the key
question is how particles are distributed over the ditTerent energy states which charac-
terize these particles. For example, if the particles move freely, we may be interested
in information about their distribution over the velocities; if the motion of the particles
is quantized, knowlcdge of their distribution over the cnergy levcls is necessary, etc.
Knowing such distributions, we may find all average characteristics of the many-particle
systcms.
6.3 Statistics 173
The rules and principles according to which particles occupy the energy states in many-
particle systems constitute the so-called physical statistics. In the physical statistics, to
describe the occupation of states by particles, one uses a distribution [1I11c/ion of par-
ticles. Under equilibrium, the distribution over the energy levels determines entirely
the properties of many-particle systems. The distrib"tion jUllctioll has the meaning
of the probability offinding particles wi/It a given energy. E. Let £, be the energy
level of particles in a many-particle system, the index I nwnbcrs the energy levels.
Then, the distribution function can be thought of as a funclion of the energy, F(£[).
Obviously.
(6.13)
where N is the total number of the particles. It turns out that the statistical principles
in classical and quantum physics are different. The quantum physics brings statistical
features that are absent in the classical description. These features are associated with the
fact that elementary particles, including electrons, are identical and it is impossible, in
principle, to sped fy their coordinates and trace a given electron. In addition, an "internal
charactcristic" ofa particle, the spin, plays a very important role in many-particle physics.
Although a definition of spin is absent in classical physics, we start with a brief review
of classical statistics.
Classical statistics
In classical physics, undcr equilibrium the distribution function has an exponential form.
This is the so-caIled Boltzmann djstribution
(6.14)
with C and T being the normalization constant and the ambient temperature, respec-
tively. The normalization constant can be found from Eq. (6.13). In general, statis-
tics in classical physics does not restrict the number of particles occupying an energy
levcl £.
As an example of the application of the BoltlJnanll distribution, consider the distribu-
tion of free particles with mass III over the velocities. First, we suppose that the system
under consideration is uniform, i.e., external forces are absent. Then, the energy of these
particles coincides with the kinetic energy £ = Ekin = mv 2 /2, where v is the particle
velocity. For three-dimensional particles, we have v = (lJx • IJ),. IJ:) and each of the veloc-
ity components can vary in the interval [-00. +00]' Thus, by rewriting Eq. (6.14) we
can introduce the distribution function as
(6.15)
Since the velocity is a continuous value, the meaning of tile distribution F u is the follow-
ing. If we define an infinitesimally small volume d 3 1J = dvx x dvy x dv: around a given
velocity v in the "velocity space," then the number of particles whose velocity is inside
this volume is F II (v)d 3 v. That is, actuallY';:1I is the density of the distribution over the
velocity. To find lhe normalization constant, we can integrate:Fl) over all possible v, and
then we obtain the lota1 number of the particles N. As a result, Ihe distribution function,
TII(v), can be written in the form
F,,(v) = N 3/2 e-",v-l/(lkRT) = N ( )m 3 /2c-m(t~+1I;+V:)/(2kflT).

( 27r k. T )
fII
27r k. T (6.16)
This is the so·called Maxwellian distribution of the particles. By using the Maxwellian
diSlribution. we can calculate average characteristics of the so-called ideal gas of parti-
cles. For example, the average energy per particle is
-E = -N1 f] mv
d v -:F:
1
2 II (v)
3 T
= -k.
2 . (6.17)
the average velocity at equilibrium is \i = 0, etc.
Fermi statistics for electrons

Now we return to the more general case of quantum statistics. As we stressed at the
beginning of this seclion, the spin of the particle plays the crucial role for quantum
statistics.
The definition of the spin ofa particle was introduced in Section 3.4 as an additional
"internal" degree of freedom. Though one can compare spin with classical rotation, in
fact, spin is strictly a quantunHnechanical quantity and differs substantially from its
classical analogue. The principal quantitative characteristic of spin is a dimensionless
quantity called the spin /lumber s. It is well established experimentally that an electron
has a spin numbcr equal to 1.
If onc fixes an axis in space. the projection of the electron
spin on this axis can be either +! -!.
or A complete description of an electron state
requires a set of quantum numbers: three of them correspond to motion of a particle
in space, say 1 = {II, 12 ./3 1. and one more corresponds to a spin, s. According to the
classification discussed in Section 3.4, this case corresponds to a two-fold degeneracy
for each energy level.
For many actual cases the electron spin is not important in altering the energy spectra,
or the spatial dependence of wave functions, etc. There is one crucially important con-
sequence of the fact that the electron spin number is a half integer. Indeed, the particles
with half-integer spin numbers obey the Pauli exclusion principle. which we introduced
in Section 3.4. It reads as follows: any quannun state (I. s) can be occupied by a sin-
gle particle only. In other words, two electrons in a system can not be simultaneollsly
in the same quantum state. Stated in another way, two electrons may be in the same
energy state if their spin quantum numbers are different (the so-called degenerate state);
if one spin quantum number is +~, the other must be -~. The degeneracy is removed
if there is interaction between electron spin and electron tr..lOslational (orbital) Illotion,
which is known as spin-orbital interactioll. In this case, the electron spin affects the
electron's spatial properties, and electrons with spin +1and -! have different energies.
In Fig. 6.3, we illustrate possible populations of the energy levels by the electrons for
two cases, degenerate and nondegenerate energy levels.
6.3 Statistics 175
(a)
6 (b)
-"""'---~ .."" .. - :=~::::;::;:

r
-' .
..........~:-- <:::::::::::..J,-__
. . _6'-__ r
--"---,"1'"- .., ._----"
.-.-
. ....
". .
r
'"
Spin degeneracy Spin splitting

Agure 6.3 Occupation of energy levels by the electrons: (a) spin-degeocralc levels and
(b) nondegeneralc (spin-split) leyels.
T1=O T1 < Tl<T3< T4
11~"':>::\\1r T,
1/2 T,
o,l-------J~=-E
Figure 6.4 lim\! the Fermi distribution function varies with cryslllllcmpcrnture.
II is clear thai the Pauli exclusion principle leads to a new, non-classical statistics of
electrons. Such statist ics is called the Fe,.",i statistics. Under equil ibrium the occupation
of the energy levels is described by the Fermi distribution jimcriol1:
I
FF(EI,s) = 1+ e(E'.J-E,)/(kBT) " (6.18)
where T is the temperature of the system, EI.s is the energy of the quantum state charac-
terized by the set of quantum numbers I and s, and E t· is the so-called Fermi energy or
Fermi l(!l'el. The evolution of the Fermi function :FF(EI.s) with the temperature is shoy:n
in Fig. 6.4, where we use the temperature paramctcrs T4 > TJ > T2 > T1 with TI = O.
Importantly, the Fermi energy can be relatcd to the total number of electrons through
the normalization condition of Eq. (6.13). Taking into account explicitly the summation
over the spin, \\1: obtain
(6.19)
which gives E. = E.(N. T).

From Eq. (6.18), one can see that in accordance with the Pauli principle the occupation
of any energy slale, {/. s I. defined by Eq. (6.18), is always less than or equal to I.
At high temperatures the second tcrm in the denominator of Eq. (6.18) is substantially
larger than unity and the Fermi distribution is close to the Boltzmarm distribution:
(6.20)
Equations (6.20) and (6.14) are the same whclllhc normalization constant C is equal to
EF-).
C=exp ( - (6.21)
kaT
TIle corresponding curve is shown schematically in Fig. 6.4 at T = T4 •
In the limit of low temperatures, T -+ O. the function FF transforms into a step
function:
:FF(E) = II.O. (6.22)
i.e., Fr(E) = I for the energy levels below the Fermi energy E~· since all levels with
E < E F are occupied and FF(E) = 0 for energies above E F since these levels are empty.
tn this limit, the electron system is frequently referred to as a highly degenerate elecflUJI
gas.
Now we can apply Fermi statistics to the electrons in a conduction band. Let II be
the concentration of electrons in the conduction band with energy dispersion £(k). We
accept that the energy spectrum is independent of the spin. Thus, the set of the quantum
numbers, I, is identical to the set of the electron wavevectors, k. According to the Fermi
distribution, the probability of finding an electron with the wavcvector k is
I
:FF(E(k» = 2, +e(E(k) E,)Ilk,T)' (6.23)
where the factor 2 comes from the spin degeneracy. The Fermi energy, E F • and the
electron concentration. ". are related through the following equation:
N 2 ,
II = II = V L: I +
k
e{E(k) £f)/(koT)' (6.24)
with V being the volume of the crystal. The summation in the latter formula can be
converted to an integration. Indeed, according to the analysis given in Section 4.4 and
Eq. (4.10), the electron wavevector takes the following values: k,L x = 2Jr/•. k),L y =
2Jr/2. and k:L: = 21f/3, with I" 12 • and 13 being integers. Here we introduced the crystal
dimensions Lx. Ly, and L:. (The crystal dimensions are related to the basis vectors of the
lattice OJ and the number of primitive cells N; as Lx = 0xNx. etc.) Thus. summation over
k is equivalent to summation over I,. The latter can be approximately calculated via the
integral: LI•.U/.')::::: f JJd/. d/ 2 dl 3 (. . .). Since the distribution fUIlction depends
on £(k), it is convenient to express the latter integral in terms of integration over k. We
can use the relationships
Lx ,.
6 II = - 6":r'
2"
6.3 Statistics 177
Since L., x L y x L:. = V, we can finally write
~( ..)= (2:)3 JJJ dk.. dky dk,( .. .). (6.25)
This procedure of replacement of the summation over djscrcte k by integration over

continuous k is useful for the calculation of average quantities.
As an example, let us calculate the Fermi energy EF of the electron system at low
temperatures (T -+ 0). The energy spectnun of the electrons is supposed to be isotropic,
i.e., it depends only on the modulus Ikl = k: E(k) = E(k) = 1;2 k 2 /(2",·), where m· is
the effective mass orthe electron. Let the electron concentration be II. According 10 the
Fermi statistics, electrons will occupy all energy stales below the Fermi energy. Since
£(k) is an increasing function of k, it follows. for T ..... O. that aLi states with k :5: h arc
occupied, where kf is the so~called Fermi wflveveclordefined through £(k F) = £~'. When
calculating the concentrntion according to Eq. (6.24), we have to perform the summation
(integration) over all occupied states, i.e., k ::s k F • For these k we have Ff = I, and it
follows that
1I=2.I>=2.
V k~tf V
x V,JJ1
(2Jf) lkl~kt·
3
d k.
Evaluation ofthis integral gives the volume ofa sphereofradius k F, i.e., 4Jfk~/3. Then, we
obtain a relationship between the Fermi wavcvector kF and the electron concentration II:
kF = (3rr 211) 1/3 . Finally, the Fermi energy of the degenerate electrons in a bulk crystal is
11 2,,2/3
E" = (3JT')'13 at T -+ O. (6.26)
2",·
In this case, E F increases as the 2/3 power of the electron concentration II. Because the
Fermi function contains the exponential factor, the low~temperature limit corresponds to
the condition EF » kB T. In metals and heavily doped semiconductors, the electron gas
remains degenerate up to room temperature. For example, in the case of a GaAs crystal
with the effective electron mass M· = O.067mo, wheremo is the mass ofthe free electron,
at the concentration 11 = 10 11 cm~3, we find kf = 1.43 X 106 cm- I and EF = 11.6 meV.
This energy corresponds to a temperature of 135 K. Thus, at T < 135 K the electron gas
concentration 11 = 10 11 cm-3 in GaAs can be considered as degenerate and it is possible
to lise the above estimates for the Fermi energy, £f. and the Fermi wavcvector, k~·.
The degenerate electron gas is an interesting and very important physical systcm. This
limiting case facilitates understanding a number ofcomplex phenomena in a simple way.
Indeed, as emphasized previously, in a degenerate electron gas all stales below £1' are
occupied. Let us imagine that a small external perturbation is applied to such a many~
electron system. The perturbation first of all will cause a redistribution of electrons
between Ihe energy states. However, below the Fermi level all states are completely filled
and no redistribution is possible. Instead, only those electrons that are at the Fermi level,
i.e., that have energy just equal to £f, can be affected by the perturbation. This results in
the fact that only a small portion of the electrons can participate in the crystal's response
to the perturbation. One can say that these "active" electrons are on dlC so~cal1cd Fermi
slIfjace in k space and the size of the Fermi surface determines the basic properties
k.
(a)
(b)
(c)
Figure 6.5 The Fermi surfaces for (a) three·, (b) 1\\'0·, and (c) one-dimensional electron gases.
of the degenerate electron gas. For bulk·like crystals with the simple energy spectrum
considered above, the Fenni surface is just a sphere of radius kF. The Fermi surface for
a three·dimensional electron gas is presented in Fig. 6.5(a). Using Eq. (2.4) it is easy to
show that an electron on the Fermi surface has the velocity lJF = tlkF/m-.
We can consider low·dimensional electron systems quite similarly. As discussed in
Section 4.5, by using heterostructurcs it is possible to fabricate artificial potential wells,
which confine the electrons from the conduction band in such narrow layers that electron
motion across the layers becomes quantized. This results in electron energies in the form
of low-dimensional subbands given by Eq. (3.49) for the so-called quantum weUs:
Ill/?
E,,("JI) = 'I, + 2111 ~, (6.27)
where 13 (/3 = 1,2, ... ) and k ll , the two-dimensional wavevector, determine the electron
mOl ion in the plane of the layer. The Fermi distribution function in the foml ofEq. (6.18)
can be used to calculate the two.Jimellsio"al electron COllcelltratioll (sheet COllcelltratio1l
ofelectro1ls), "20:
(6.28)
where we have introduced the sheet concentration of electrons. n20 (the number of elec·
trons per unit area), the thickness d, and the area S of the confined layer; obviously
V = d x S. Actually, exactly as found for bulk crystal. Eq. (6.28) establishes the rela-
tionship between the sheet concentration, n2D. and the Fermi energy, £F .
6.3 Statistics 179
The sheet concentration of the e1ectrolls can be explicitly calculated in the limit of
a degenerate electron gas. Let us suppose that the temperature is low and that only the
lowest subband is populated by the electrons. Then, in the sum over I J in Eq. (6.28) we
should keep only one tcrm with /3 = I:
2 I
IIZD = S L I + e(£llkll ) £F)/(kllT)' (6.29)
'"
Calculation of the righI-hand side of this equation can be simplified by replacing the
summation over k ll by integration, similarly to thai of Eq. (6.25):
l).)=
,
S,
(2rr)-
!!dk.,dk y ( ...). (6.30)
where kx and k.l' arc the components of the two-dimensional vector kilo In the limit of
T -+ 0, the sheet concentration 112D can be calculated by defining a two-dimensional
Fermi wavevector kll.F via the relationship
t; 2k 2
E, = 11.' (6.31)
2m"
The Fermi "surface" for two-dimensional carriers is a "disk" of radius kll .F in k ll space,
as shown in Fig. 6.S(b). On performing the integration ofEq. (6.30) over the "disk," wc
find
and (6.32)
That is, the Fermi energy, E F , increases as the first power of thc clectron sheet concen-
tration, "20.
For a nanostructure in which the electron motion is restrictcd in two directions, i.e.,
for thc quantum wire, the encrgy spectrum is givcn by Eq. (3.50):
t; 2k 2
E'l.!)(k.T ) = Etl.!) + 2 :'
11/
(6.33)
where it is assumed Ihat the x direction is the only direction of free motion. By applying
a procedure similar to that used above, wc may sec that for low temperatures the Fermi
"surface" shrinks to two points in the one-dimensional k space: kx = ±k F with kF =
~JTnlO' Here, /110 is the linear electron concentration, which can be defined by the
formula 11 10 = N / L, with N being the total number of electrons in the wire and L
being the wire length. Subsequent evolution of the Fermi surface of the electrons as
the dimensionality is lowered is illustrated in Figs. 6.5(a)-{c). The Fermi energy of
one-dimensional electrons is givcn by
JT 2t;2
EF = ll~o' for T ~ O. (6.34)
8m"
On comparing the Fermi energies obtained for djfferent dimensionalilies oflhe electron
gas, we can conclude that lowering the dimensionality of the gas gives risc to a more
rapid increase of E F with increasing conccntration for low-dimensional systems.
The concept of Fermi statistics is one of the fundamental ideas of modern solid-state
physics and extremely important for nanoclcctronics. The Fermi statistics is applied
widely in the description ofnanoclcctTOnic devices.
6.4 The density of states of electrons in nanostructures
To complete the analysis of the occupation of the energy levels by electrons in nanos-
.ruelures. we shall study another physical quantity characterizing the occupation, known
as lhe del1sityo!srates. The reason for introducing this quantity is the following. From
the results presented in the previous chapters, we see that the electron energy spectra
in nanostructures with conduction electrons are complex and consist of a series of sub-
bands. The distances between subbands are determined by the profile of the confining
potential, while inside each subband the spectrum is continuous and these continuous
spectra overlap. To characterize these complex spectra, it is convenient 10 introduce a
special function known as the density of states, e(E), which gives the number ofquantum
states dN(E) in a small interval dE around energy E:
dN = e(E)dE. (6.35)
If the set of quantum numbers corresponding to a certain quantum state is designated as

v, the general expression for the density of states is defined by
erE) = L 8(E - E,). (6.36)

"
where E", is the energy associated with the quantum state v. At this point, it is useful to
introduce Dirac's famous a-function:
for x #0.
8(x) = 1°'
00, for x ~ O.
(6.37)
though
f~~ 8(x)dt = I. (6.38)
Dirac's a-function is used when performing an integration. The main rule of such an
integration is the following:
l' dx 8(x - xo)(x) = (xo). if a < Xo < b. (6.39)
where et>(x) is an arbitrary well-behaved function.

As a simple example ora calculation of the density or states by using the definition
of Eq. (6.36), let us calculate this quantity for electrons in a bulk crystal. As discllssed
previously, for such a case the set of quantum numbers is {k. s}. Assuming the energy
to be independent of the spin, E(k) = tilk 2 /(211/·). Replacing the summation by an
integration as in Eq. (6.25), we obtain
l?JO(E) = 2 V)
(2" )
fff dk8(E - E(k»). (6.40)
6.4 The density of stales 181
{)(E) (a) {)(E) (b) •
•
• - -'
•
E o" " '3 E
(e) (d)
{)(E) ()(E)
\j\J\ ,• ,,
0 0
'II " '" E Ell1
"'''' F.' E
Agure 6.6 The density of states of electrons, 'l(E). in systems of different dimensionalities: (a) a
bulk crystal. (b) a quantum well. (e) a quantum wire. and (d) a quanlum dot. Here El. Ell. and
EIlI are the ground stales in a quantum well. quantum wire. and quantum dol. respectively; £2
and £} arc higher states in a quantum well and Ea, E b • and E( are higher stales in a quantum wire
and in a quantum dot.
Because E(k) actually depends on the modulus It, spherical coordinates may be used.
Integration over two angles gives 41l' and the triple integral is reduced to the single
integral
e31J(E) = 2V 3 X 4" 1~ dk "'8(E - E(k)). (6.41)

(2") _~
Now, we replace the integration over k by an integration over £(k) = E. Then, taking
into acconntthat k' = (2m" /1,')£ and dk = jm" /(2h'£)d£,
- , )J/'l~ d£ Jl8(E - £).

e3D(E) = -V, (2m" (6.42)
2iT- 11- 0
Finally, by using the rule ofEq. (6.39), we obtain

",. )3/2 -hE.
V
e3D(E) = ( tl2
-
IT 2
(6.43)
The obtained density of states for threc·dimensional electrons is depicted in Fig. 6.6(a).
For electrons in a quantum well with the energy spectrum given by Eq. (6.27), the
set of quantum numbers includes a spin quantum number, S, a quantum number, 13 ,
characterizing the transverse quantization of the electron slales, and a continuous two-
dimensional vector, k.jl' Hence, v = {s./ 3 • kU}' There is a two-fold spin degeneracy of
each state, (s = ± ~), so that
(6.44)
In order to calculate the sum over kx and k'l" we can apply the replacement, Eq. (6.30),
with S being the area of the surface of the quantwn well, S = Lx X L" where Lx and
,
L y are the sizes of the quantum well in the x and y directions, respectively. Calculations
of the integrals give liS
Qm(E) =
",+S
h' ' "
1,:(ldt: ,(E - <I, - t:) =
m·S
fl' ' " 8(E - <I,).
Jr-~o Jrl·~
/) I)
(6.45)
where 8(x) is the Heaviside step-function:
8(x) = II.O. for x > O.

for x < o.
(6.46)
Very often the density of states per unit area, (}2DI S, is used to eliminate the size of
a sample. Each term in the slim of Eq. (6.45) corresponds to the contribution from one
subband. The contributions of all subbands are equal and independent of energy. As
a result, the density of states of two-dimensional electrons exhibits a staircase-shaped
energy dependence with each step being associated with onc of the energy states, Ell'
The height of each step is universal and depends only on the effective electron mass.
Figure 6.6(b) depicts the two-dimensional density of states. By comparing the densities
of states for the electrons in bulk crystals per unit volume and those of quantum wells
per unit surface we may sec that the differences between the two- and three-dimensional
cases are most pronounced in the energy regions of the lowest subbands. For large /3
the staircase function lies very close to the bulk curve 1230(£) and coincides with it
asymptotically.
Similarly, we can find the density of statcs of a one-dimensional electron gas with the
energy spectrum given by Eq. (6.33). The result of calculations is
QID(E) = L QI,.I,(E).
h.lJ
where
(6.47)
Here, L is the length of the wire. Schematically, 1210(£) for one-dimensional electrons is
shown in Fig. 6.6(c). The characteristic feature of the one-dimensional density of states
is its divergence ncar the bottom of each of the one-dimensional subbands. The density
of states then decreases as the kinetic energy increases. This bchavior is very remarkable
because it leads to a wholc class ofncw electrical and optical effects peculiar to quantwn
wires.
Now, we can consider the ultimate case of the density of states for zero-dimensional
electrons, i.e., for electrons in quantum dots. According to the definition ofEq. (6.36),
in the case of quantum dots or boxes the spectra are discrete. Thus, the density of states
is simply a set of t5-shaped peaks, as depicted in Fig. 6.6{d). For an idealized system, the
peaks are very narrow and infinitely high, as illustrated in Fig. 6.6(d).ln fact, interactions
bctwecn electrons and impurities as well as collisions with lattice vibrations bring about
a broadenjng of the discrete levels and, as a result, the peaks for physically realizable
systems have finite amplitudes and widths. Nevertheless, the major trend of sharpening
6.5 Electron transport in nanostructures 183
of the spectral density dependences as a result of lowering the dimensionality of the

system is a dominant effect for nearly perfect structures at low temperatures.
The dramatic changes in the electron density of states caused by dimensionally con-
fining crystals manifest themselves in a variety of major modifications in conductivity,
optical properties. etc. Indeed, as we will see, these modifications in the dcnsity of states
also lead to new physical phenomena.
6.5 Electron transport in nanostructures
As indicated in Section 6.2, there are distinct regimes of electron motion in nanostruC M
tures. In this section, we will study several instructive examples of electron transport
induced by an electric bias. We start with the simplest low-field dissipative classi-
cal electron transport, which is realized at large interconlact distances, according to
Eqs. (6.10) and (6. I2). This transport regime can occur in bulk samples, quantum wells,
and quantum-wire structures.
Classical dissipative transport

In a solid the electrons are in constant motion. However, this motion is chaotic as a result
of random scattering by imperfections. lattice vibrations, interface rouglmcss, etc. As a
result there is no preferred direction for electron motion. Thus, the net electron flux and
electric current arc equal to zero. Ifan electric field F is applied to the solid, an electric
force, -eF, acts on each of the electrons (here we suppose that the electron charge is
equal to -e). Though the chaotic character of electron motion can remain, a directional
net drift of electrons is induced by the electric force. Figure 6.7 demonstrates chaotic
electron motion without and with the electric field applied. To describe this motion, we
can use the Newton equation (2.8) for the average velocity of the electrons, v. To take
into account the scattering processes, which lead to the loss in the directed componenl
of the velocity, we introduce an additional term that contains a friction force:
• dv ",.
//I - = --v-eF, (6.48)
dt tc
where Tc can be interpreted as the momentum relaxation, or free-flight time, which was
discussed in Section 6.2. Obviously, the electron motion described by the above equation
is dissipative in nature.
For the stationary state, dv jdt = 0, we obtain
eT,
v = - -m·F = -iLF.- (6.49)
where we have introduced the parameter I.l, called the electron mobility. The mobility
is one of the basic characteristics of electron transport in the case of low electTic fields.
The definition of J.L.
eT,
iL = - . (6.50)
m-
Total displacement of the Directional displacements caused

electron between collisions by the electric field force, -eF
Electron motion between collisions

(a) due to thennal velocity of electrons (b)
Agure 6.7 ChaOlic diffusion morion of an electron OIl equilibrium (a) and when an electric field F
is applied (b).
shows that the mobility is larger for materials with small effective mass m* and with
suppressed scattering, i.e., for large reo The negative sign in Eq. (6.49) reflects the fact
that the electrons move in the direction opposite to the elcctric field, because of their
negative charge.
Having the electron average (drift) velocity, v, and the electron concentration, 11, we
can calculate the electric current density as
J = -e\'ll = e/1-nF = erF. (6.51)
Here,
er = eJ.t" (6.52)
is the so·called specific conduct;v;ty, and, as we can sec, it depends on both the electron
concentration, ", and the electron mobility. J.t. The result given by Eq. (6.51) is known
s
as Ohm law. If the specific conductivity and the geometrical dimensions of a sample
arc known, we can readily calculate the tala I electric current:
<1>0
I=J·S=<1SF. III = - . (6.53)
R
where we introduce the cross-section of the sample S. the voltage drop across the sample
<1>0 = L.r . F, and the sample length L.r. The resistance R is
L,
R=<1S. (6.54)
Equations (6A9H6.51) are valid for systems of any dimensionality, where dissipa-
tive transport occurs. These include three-dimensional, two-dimensional, and one-
dimensional systems. For all of them, the electron mobility is a representative char-
acteristic of the classical transport regime.
To define measurement units of tile mobility, we must first discuss briefly lhe units of
measurement of several entities introduced in the preceding discussion. (1) The charge
is measured in units of coulombs, C. The elementary charge of the electron, e, is equal
to 1.6 x 10- 19 C. (2) The electron concentration is measured in m- 3 , m- 2 • and m- l for
three-dimensional, two-dimensional, and one-dimensional systems, respectively. (3) The
electric field is measured in units of volts per meter, V m-I. (4) Accordingly, the units
of mobility are 111 2 V-I 5- 1 . (5) The current is measured in amperes, A, i.e., coulombs
per second, C s-J . (6) For the current density and conductivity of Eq. (6.51). we obtain
different units for electron systems with different dimensionalities. For example, in a
bulk crystallhe current density is measured in amperes per unit cross-section, A m~2,
and the units of conductivity are I Q-l m- l with Q being the ohm-the unit of electrical
resistance.
Returning to the discussion of mobility, we shall stress that the magnitude of the
mobility depends 011 the particular material and scattering mechanisms in this material.
Typically, the scattering mechanisms in semiconductors include scattering by impuri-
ties, imperfections, and lattice vibrations. If the first two scattering mechanisms can
be avoided in pure and high~quality samples, lattice or, as one frequently says, phonon
scattering can not be avoided in principle. The total scattering rate is a sum of rates of
scattering due to particular mechanisms. Each of the rates is proportional to a scattering
probability and inversely proportional to the mean free time. This leads to the conclu-
sion that also the mobilities due to two or more scattering mechanisms should be added
inversely:
I I 1
-=-+-+ .... (6.55)
p., Jiph .aim
where J.Lph. J.Lim, ... are partial mobilities determined by phonon scattering, impurity
scattering, etc. Phonon scattering depends in an obvious way on temperahlrc: at low
temperatures, when thermal lattice vibrations are suppressed, the rate of this type of
scattering is small, but it increases with the temperature. This results in a finite value of
mobility even in pure crystals. For example, at room temperature the electron mobilities
of Si and GaAs bulk crystals are limited to 1350 cm 2 V-I s-I and 8500 cm 2 V-I s-I,
respectively. TIle same order ofmagnitude for J.L is realized in low~dimensional systems at
room temperature. When the temperature decreases, mobility increases and the scattering
by imperfections and impurities becomes the limiting factor, as shown in Fig. 6.8. These
scattering mechanisms can be avoided in low-dimensional systems, where mobility can
reach values above 105_10 6 cm 2 V-I S-I.
The electron current in the form of Eq. (6.51) is valid for uniform conductors. I f the
electron concentration, II, is spatially dependent, the electrons, naturally, diffuse from a
region of high concentration to one of a low concentration. This produces an electron
flux opposite to the gradient of the electron concentration: --dll/dr. The diffusive
contribution to the current can be written as
dn
Jo =eD- =eDVrn. (6.56)
dr
10" r--"":7'..--.,=--;--r------,
10 cm-
14 3
-'""'
I Lanice
-'" scaucring scattering

In T
N
>- 163
E
<.)
~
,,: 1
"
50.~~----,~~--=-_____;')
100 200 500 1000
T(K)
Rgure 6.8 The temperature dependence oflhe electron mobility in Si for a system with two
mechanisms of scallering., lattice scanering and impurity scattering. The concentration of
impurities, No. is indicated near each curve. After W. F. Beadle, 1. C. C. Tsai el al. (Eds.). Quick
Reference MOII/wljor SemicOlldllctor Engineers (New York. Wiley. 1985).
where D is the diffusion coefficient, that was discussed in Section 6.2. Thus, in a nonuni-
form conductor, the total current is composed of both drift and diffusive contributions:
(6.57)
It is easy to establish the fundamental relationship between two kinetic coefficients,

the mobility, 1-'-, and the diffusivity. D. Indeed, let us apply Eq. (6.57) to equilibrium
conditions. Under equilibrium, the electric current is absent, i.e., J = O. The electron
concentration, nCr), can be expressed via the electrostatic potential, <%>(r), according
to Boltzmann's distribution in the form of Eq. (6.14): nCr) = "0 exp[e<l>(r)J(kBT)]. On
putting this dependence and the electric field, F(r) = -d<l>Jdr, into Eq. (6.57) and equat-
ing the current to zero, we obtain the so-called Einstein relationship:
D k.T
-=- (6.58)
" e
Thus, having the electron mobility I-L, one can easily calculate the diffusion coefficient,
D.
The results just discussed are relevant to steady-state electron transport. However,
the ewton equation (6.48) can describe the behavior of electrons in an arbitrary time-
dependent electric field, F(I). Since any dependence, F(I), can be represented by using
a Fourier transform, without loss of generality we can analyze the case of a harmonic
dependence of the field:
F(f) = F~COs(Wf). (6.59)

where Fw is the field magnitude and w is the frequency of field oscillations. For further
analysis, we use the well-known formula cos(wr) = ne[e-ian ]. Here ne[.. .) means
calculation of the real part of an expression in the brackets. Thus, the physical field is
F(t) = ne[Fwe-;""j.
It is worth noting that calculations with exponential functions like e- iwl arc always
simpler than those with sine and cosine functions. For this reason, the following approach
is commonly used. Instead of the field in the form ofEq. (6.59) in the Newton equation,
one uses the complex representation of the field:
F(t) = F",e -iWl .
When a solution to Eq. (6.48) with the complex field is found, the real part of this solution,
that has a physical meaning, can be calculated easily.
To apply this approach, we look for the solution of Eq. (6.48) with a complex field of
the exponential form vwe- iw,. On substituting this form into Eq. (6A8), we immediately
obtain
e r,
v'" = - - . F<t,. (6.60)
III· 1- twte
In this case, the electron velocity should be calculated as Vel) = 'R.e[vwe- iw1 ]; i.e.,
ert (COS(wt) Sin(WI))
v(r) = - - F w
•
+ wrt ,. (6.61)
m* 1+ w 2r 2 , 1+ Wlr-,
Thus, the cosine electric field of Eq. (6.59) induces electron motion, with both cosine
and sine contributions. It is instructive to rewrite Eq. (6.61) as
v(r) = _ er, F eos(wr - 11') (6.62)

w
m* JI + w 2r 2 t
where we introduce the phase shift rp of the electron velocity with respect to the phase
of the field (6.59). This phase shift can be found from the equation tan 1P = WT t . Here,
qJ > 0, which implies a delay of the electrons with respect to the field changes. Obviously,
the delay exists only for non-zero frequencies wand arises because of the friction force
in the Newton equation. The second conclusion, which follows from Eq. (6.62), is that
the magnitude of velocity oscillations decreases when the field frequency increases. In
the limit of very high frequencies WTc » I, the alternating electron velocity vanishes.
Having the electron velocity, vet), and concentration, II, we can calculate the time-
dependent electric current similarly to Eqs. (6.51) and (6.52). However, more common
is the use of the complex representation of the current density:
(6.63)
where we introduce the complex COlldlictivity:
cr(w) =
e'r,1I
m*
(1+
1
,
w-r c-
,+ I .+ I
wr,
")
w-r;
,
)
. (6.64)
This conductivity reduces to its steady-state value in the limit WTc -4 0, and in this
limit, Im[cr] --+- O. At finite frequencies. both contributions, 'R.e[a] and Im[cr], are
0.8
I
0.6
a(W,,)
<J(O) 04
0.2 2
0.00 1 2 3 4 5 6 7 8 9 10
W"
Agure 6.9 Dependences of Ihe real (curve I) and imaginary (curve 2) pam of the complex
conductivity of Eq. (6.64) on frequency w; u(O) = a(w = 0).
important for an alternating current; in particular, at w = I/Te. the two contributions

are equal muncrically. The latter property is used typically to determine t'c by varying
the frequency, w. Importantly, the electron scattering time, Tet determines the high-
frequency properties of a material. In particular, at W!c » I the conductivity vanishes
and the electron subsystem of the material docs not react to a high-frequency field.
The relationship given by Eq. (6.64) is known as the Drude formula. In Fig. 6.9 the
dependences of'Re[a(w)] and Im[a(w)] versus w arc shown.
The conductivity O'(w) is a specific characteristic of a material. If the material is
homogeneous and its geometrical parameters are known, we can use a to relate the
total alternating current i and the alternating voltage d> = 4>we-ic.>I, both in complex
representation:
Lx
Z(w) = u(w)S (6.65)
Here, in the second relation, we omit the vector designation; Lx and S are the intercontact
distance and the cross-section of the sample, respectively. By comparing these results
with those of the steady-state case given by Eqs. (6.53) and (6.54), one may see that,
instead of the resistance, R, a frequency-dependent parameter, Z(w), is introduced. II
is called the impedance. The impedance is a complex function Ihat characterizes the
electrical properties of the whole sample. The impedance can be introduced for any
nonhomogeneolls sample and device.
Dissipative transport in short structures

The mobility is a specific characteristic of a uniform conducting material. In a sample
of extended djmcnsion L.n the electric field is almost uniform. It can be estimated as
F = 4>0/ Lx, with <1>0 being the applied voltage, and the mobility determines basically
(a) .
n+ I n+
(b) Vex)
E,
x
(e) Vex)
v = -eo
Vex) - -x 3/2
Figure 6.10 (a) The structure schematic of an n+ - i - n+ diode. Potential energy profiles, V (x),
in a diode with charge·limilcd dissipative transport: (b) for an unbiased diode and (e) for a
biased diode.
the electric rcsistcUlce and the current, according to Eq. (6.53). In short samples, another
electrical effect can contribute essentially toward the electron transport and the current.
The effect arises due to nonuniform redistribution of the electrons and, thus, of the
electric charge across the short sample. The charge induced by the current affects the
potential distribution along the sample, making it strongly nonuniform and dependent
on the current. As a result, the dependence of the total current on the applied voltage
is no longer linear. The electron transport in this regime is called space-chal'ge-limited
trallsp0rt. Since in this text we are interested particularly in very short structures, it is
instructive to review briefly the space-charge-limited transport under the condition of
Eq. (6.1), when we can introduce mobility.
Let us consider a short sample with two contacts as depicted in Fig. 6.2(b). The two-
contact device is called the diode and the body of the sample between the contacts is
called the base of the diode. Assume for simplicity that the base is not doped, i.e., there
are no conducting electrons in the base. The contacts can be fabricated by heavy n-type
doping of the contact regions; they are called n+ regions. In such a case, this device is an
n+ -i (insulator)-n+ diode as shown in Fig. 6.1 O(a). Let 4> designate the electrostatic
potential. Then, the potential energy of the electrons is V = -e<P. In Fig. 6.IO(b), a
sketch of the potential energy of an unbiased diode is presented: the electrons in the
contact regions on the left and right are separated by a high potential barrier in the base.
Ifan electric voltage <Po is applied, the potential relief changes, as shown in Fig. 6.1 O(c).
The potential energy. V(x) = -e<P(x), decreases and some ofthe electrons can overcome
the potential barrier. penetrate from the emitter electrode (the cathode) into the diode
base, and contribute to the current. This bias-induced effect is referred to as electron
injection from the electrode to the base. Obviously, with increasing bias, the maximum
of the potential shifts toward the cathode and becomes lower, and the injection current
increases. nlis physical picture can be described by the following simple model, which
is valid at large electric biases. The current density, J. can be written via the mobility
characteristic for the undopcd base, J.L. the injected electron concentration, n(x), and the
electrostatic field, F(x) = -d<J>/d\", as in Eq. (6.51):
J = eIJll(x)F(x). (6.66)
Importantly, for the chosen configuration, see Fig. 6.1 O(a), the cathode is on the left and
the anode is on the right. The potential, $, increases with x. and the electric field, F,
and current density, J, are both negative. As a result of the continuity condition, in the
stationary state the current density through the diode is conserved: J = -)0 = constant,
whcre )0 is the absolute value of the current dcnsity. From the latter relationship, we can
determine the concentration of the injected electrons, n(x):
II(X) = -Jo/(eJ.LF(x)). (6.67)
For the electrostatic field, we can use Poisson's equation,

d' dF ell(x)
(6.68)
dt 1 = - dt = E"oE
where € is the dielectric constant of the base material and €o is the permittivity of free
space. After substitution of n(x) from Eq. (6.67), Eq. (6.68) can be rewritten in terms of
the field, F:
or (6.69)
This equation should be supplemented by a boundary condition for the field. We can use
the fact that at the maximum of the potential barrier:
d<l>
- dT = F = O. (6.70)
At a large electric bias, this maximum is shifted close to the cathode and we can set
F(x = 0) :::::: O. This approach neglects the processes which occur in a very narrow
region between the real cathode and the potential maximulll. It is called the virl/lal·
cathode approximation and it is used widely for such a simplified analysis. Integration
ofEq. (6.69) leads to the solution
I, )0 20l0 )'/' X 1/2.

-F-(x) = x or F(x) = _ (6.71 )
2 €0€/.L
( €O€/.L
where the negative value of the square root is chosen for the solution in accordance
with the above-discussed sign of the electric field. Thus, the distribution of the potcntial,
(x), is given by
(x) = _
1"o
dT F(x) = ( 80l0
9E"0E" I.l
)'/2 X 3/ 2 (6.72)
From the total voltage drop <1>(L~) = <1>0, we can find the relationship between the
current density, lo, and the bias, <1>0. i.e., the current-voltage characteristic:
9£Ofj1 <1>~
Jo = JO.d = 8 V· (6.73)
.,
This is the so-called MOll-Gurney law for the diode with dissipative electron trans-
port. We have found that the current-voltage characteristic becomes strongly nonlinear
because of the space-charge effect. The space-charge effect can be characterized, partic-
ularly, by the average concentration of the injected electrons, Ii:
n = -
I
L~
1 0
L
'
n(x)dr = -3 £0£<1>0
2 eL;
,. (6.74)
which increases with the applied voltage.

Equation (6.49) and F(x) facilitate the calculation of the average time of electron
transit through the diode, 'tr.d:
I tr d =
. 1
0
IlL.
L
'
d\'" - - =
u(x) 0
d\'"
1
IlF(x)
=
4
-/0 d'
3·
(6.75)
where
L'
10d = ~ (6.76)
. Il¢>o
Here. 10.d is the transit time of the electrons drifting in the average electric field, <1>0/ L~,
i.e., when the space~charge effects arc neglected. As follows from Eq. (6.75), these effects
increase the transit time by the factor 4/3.
In general, non-stationary electrical properties of a device are characterized by the
current induced in response to a time-dependent external voltage bias. If the bias is ofa
frequency w, the current response is given by the impedance according to Eq. (6.65). For
a Mott-Gurney diode biased with a steady-state voltage, <1>0, the complex impedance
can be calculated exactly:
Z(w) = 6R
11 3d [ (11 - sin 11) 2"
+ i (11' - 1 + cos 11 )J . (6.77)
where
d<1>o 4L;
Rd = - = ::-----c:-:-
dlo 9€o€J1S¢l0
is the differential resistance of the diode at steady state calculated from Eq. (6.73),
10 = loS. S = L.l' X L~ is the cross-section of the sample, and Q = Wl tT.d; i.e., it is
determined via the average time,/tr.d, ofelectron transit through the diode; see Eq. (6.75).
The dependences ofRe(Z(Q)] and Im(Z(Q)] are presented in Fig. 6.11. The impedance
of Eq. (6.77) differs considerably from the impedance of a material with the Drude-
like high.frequency behavior. Importantly, as the voltage bias increases, the impedance
decreases and the current response increases ex: <1>0; moreover, the spectral width of the
response becomes larger (ex: 1/ IIT.d <X <1>0). Both conclusions provide evidence that a short
08
06
2(wt".d)
2(0) 0.4 2
0.2
000 I 2 3 4 5 6 7 8 9 10
Wllr.d
Agure 6.11 Dependences of the real (curve 1) and imaginary (curve 2) parts of the impedance of
Eq. (6.77) on frequency w; 2(0) = Z(w = 0).
diode with space-charge-limited dissipative transport under a high bias can operate al
higher frequencies than can a doped diode with Drude conductivity.
To conclude this analysis of dissipative electron transport in short samples, we wanl
to point out that the main results obtained are valid at large biases, when the diffusive
contribution to the current can be neglected:
dll(x)
/-LIF(x)l» D . (6.78)
dx
(A related analysis is proposed in Problem 7.) For such conditions, the transport is
determined 10 a large extent by the space-charge effects, the electric field is highly
nonuniform, and the injection current increases quadratically wilh the bias, as Eq. (6.73)
shows. Under high bias, lhe transit lime through the diode decreases and lhe device
remains electrically active in a frequency r:mge enlarged proportionally with the bias.
Hot electrons
Now, we return to large uniform crystals to study the effect of high electric fields on
the drift velocity and the current-voltage characteristics. Equations (6.49)-{6.51) were
obtained under the assumption that the relaxation lime, Te , is constant; i.e., it does not
depend on the magnitude of the electric field. When the field increases, the electron gas
is far from equilibrium. In particular, the average electron energy increases. This may
be understood from the following qualitative considerations. The electron energy in an
electric field may be described through the relationship
dE E- E
-=~F)- ~. ~7~
dr TE
The first term on t.he right-hand side corresponds to the power gained by the electron
from the electric field; the second term represents the rale of electron-energy losses.
The losses are proportional to the deviation of electron energy, E from its equilibrium
I
value, E cq , and TE is the energy relaxation time introduced in Section 6.2. The electron
energy increases in the electric field until the overall energy balance is maintained. In
the stationary case, when dE/dl = 0, Eq. (6.79) yields an electron energy, E. given by
(6.80)
According to Eq. (6.49), the velocity, v, is linearly proportional 10 the electric field, F.
Thus, the mean electron energy increases as the square of the electric field and
exceeds the equilibrium value, £r:q. It is convenient - and conventional in semiconductor
electronics - to consider the electron effective temperature, Te , instead of the mean
electron energy. The relationship between the temperature and the mean energy can be
found at thermal equilibrium and is £ = ak B Te /2, where the factor a is the dimension-
ality of the structure. Obviously. under thermal equilibrium the electron temperature, Te.
coincides with the lattice temperature, T. Under nonequilibrium conditions the t\\'o tem-
peratures, T. and, Te • may differ. The effective electron temperature expressed through
the mean electron energy serves as a gauge of the nonequilibrium state. If Tc exceeds T
only slightly. and the electron transport still obeys Ohm's law, we have W(lrm electrons.
The case with Tc » T corresponds to the situation in which the electrons are far from
equilibrium. For such a situation, the electrons frequently are called hot electrons. The
electron temperature can reach magnitudes of abollt several thousand degrees Kelvin,
while the lattice can remain cold. For simple estimates it is accepted that transition from
the warm-electron regime to the hot-electron regime occurs at the electric field F = Fhc.
when e(vF)T£ is equal to E eq = ak o r. It is easy to estimate the heating electric field
Fhc ' Indeed, in Eq. (6.80) the drift velocity v can be estimated from v = -ILF; thus, the
critical heating field is
Filc =
jctkBT . (6.81)
elL TE
For hot electrons, the scattering processes themselves become dependent on the field.
The linear relationship of Ohm's law of Eq. (6.51) is no longer valid and the current-
voltage characteristics, J = J(F), and the drift-velocity-field dependence, u = v(F),
can exhibit strongly nonlinear behavior, which depends on the electron bandstructure
and specific scattering mechanisms. In particular, these dependences are of different
shapes in the two mosl important materials. In group IV semiconductor materials, the
J(F) and v(F) dependences lead to the saturation effect at large fields, as shown in
Fig. 6.12(a), whereas in III-V compounds, after a nonlinear portion with a rise, the
current and the velocity undergo a decrease in some interval of the fields, as shown
in Fig. 6.12(b). As follows from the analysis of different transport regimes given in
Section 6.2, for high-speed devices the important material parameter is the maximum
of the drift velocity, which can be achieved in high electric fields. A comparison of the
saturated and maximum velocities for reI event materials is presented in Table 6.2.
One can see that somc ofthc II I-V compounds (including GaAs, luP, and InSb) have
velocities several rimes larger than those ofSi, GaP, and AlAs. For Si the drift velocity is
restricted to a magnitude of about 101 cm s-l that can be reached for electric fields above
Table 6.2 values of the saturated (for

SI, SiC, SiC" AlAs, and Gap) and
maximum (for GaAs, Inp, InAs, and
InSb) drift velocities for semiconductor
materials at room temperature
Velocity
Material (10 7 ems-I)
Si I
SiC 2
SiOl 1.9
AlAs 0.65
GaP LI
GaAs 2
InP 25
InAs 4.4
InSb 65
(aj (b)
lOJ r--~----:;'" "'=-.., 107
-
'1'
E
Si
d'~,o I" 2
u ~ E
..: 102 106 E u
"./ /cI o
~
-
0
...,
~
77 K/ /
/ ' / ' /'--FOlincs
,
"'- -
'>1 GaAs
'193 K. -FI!21mcs
./ .198 K' -.Low-Iidd c:xtmpo13110n
10 10' OL-~~~~~..---J
10 102 10J 1()4 o 5 10 15 20 25 30 35

F (V cm-I) F (kV CI11- I )
FlgUrt 6.12 Currcnt-vollagc characteristics al large electric fields; (a) Si and (b) GaAs and InP.
Reprinted with permission, (a), from E. J. Ryder, "Mobility arholes and electrons in high
electric fields," Ph)'s. Rev., 90. 766 (1953). () 1953 by the American Physical Society.
several kV em-I. For GaAs. the velocity maximum of2 x 101 em S-I can be achieved
at fields ~3.S kV em-I.
Transient overshoot effects

Limitations in the drift-velocity characteristic for a particular material can be overcome
by utilizing another hot-electron effect known as \'t!locityovershoot. In order to explain
this phenomenon, let us mention that our previous conclusions have been made for
the steady-state case, in which the electron distribution is stationary. In other words,
the previous analysis has referred to the electron properties averaged over times much
greater than the characteristic times of the system such as the mean-free-flight time,
(a) (b)
6
5 GaAs
I
20kVcm- 1
(\\OkVcm- 1
~ 1.0
<:: kVcm-
in 4
\
u E3
~
~
u 5 kVcm- 1
o
-,.
'-'0.5
~
S 2
~
,. 1 I kVcm- 1
'/
0.00 1 2 3 4 5 6 7 8 9 10 1 2 3 4
Time (10- 13 s) Time (10- 12 s)
Figure 6.13 The overshoot effect - transient response of the electron drift velocity 10 step·like
pulses of the electric field at room temperature. The field magnitudes arc indicated for each plot:
(a) Si and (b) GaAs. Reprinted with permission, from V. Mitin, Y. Kochelap, and M. Stroscio,
QUlInllim Heteroslrllc/ures, Fig. 7.13 (Cambridge, Cambridge University Press, 1999).
momentum and energy rcla,<ation times, ctc. Now, we will consider processes that occur
in the electron system immediately following its deviation from the equilibrium state. In
this particular case. we will focus on the response to the electric field in the form of a
step-function.
In general, the momentum relaxation time, t e , entering Eq. (6.49) is shorter than
the eneq,J)' relaxation time, rE, which determines the electron energy; see Eq. (6.79).
Therefore, the velocity response 10 Ihe electric field step is faster than the energy
response for the case described by Eq. (6.79). Typically, if the scattering rate increases
wilh increasing electron energy, Ihe electron velocity may exceed the stationary veloc·
ity during a time interval of the order of rEo In other words, the transient velocity
is not just a function of the electric field but also a function of the electron energy.
Indeed, the velocity "adjusts" itself quickly to the quite slowly changing energy, and
it "follows" that energy until the energy reaches the steady state. Initially, when the
electron energy has not reached the stationary value, the electron velocity correspond·
ing to the transient energy is higher than the velocity corresponding to the stationary
energy.
The overshoot effect is shown in Fig. 6.13 for Si and GaAs. The drift velocity is
presented as a function of time for several electric fields. The electric field is assumed to
be switched on at time t = O. The overshoot effect is pronounced in high electric fields.
The maximum of the transient velocity can exceed the stationary saturated velocity by
as much as two to four limes.
From a physical explanation of the overshoot effect, one can understand how it is
possible 10 utilize this effect. Let us imagine that cold electrons enter an active region
ofa semiconductor device through a contact. If there is a high electric field in the active
region, the electrons will be accelerated. At some distance from the injecting contact the
(a) (b)
--
~15
20
ND ,'" - --
1.0
'i'
6
/,V~_
0.5
0.4
"
u ,/11 E
u
~
\
E
0.3 ;;-
> 10 0.5 "
-" -
0
/ 0.2 ~
~
k. 5
~
;£ /
/
\.\ , 0.1 '"
0
0.0 0.5 1.0
0.0 - "
o - -'
0.0
/
0.5 1.0
0.0
x(~m) x (~m)
Agure6.14 Spacc-chargc-limilcd tTunsporl with the velocity overshoot effect in an n+-i-n+

GaAs diode at L~ = 0.37 11m, T = 77 K, and IP o = 0.5 V. (a) The doped regions, the electric
fields, F, and the injected electron concentration, II. (b) The drift velocity. u, and the avemgc
energy or the injected electrons, E. From A. Ghis, E. Constant, and B. Boilliaux, "Ballistic and
overshoot electron transport in bulk semiconductors and in submicronic devices," J. Appl. Phys.,
54,214-221 (1982). Reprinted wlh permission from A Ohis, E Constant, and B. Boittiaux,
Journal ofApplied Physics. 54, 214 (1393). 01983 American Institute of Physics.
electrons will reach the maximum overshoot velocity and after that their velocity will
gradually decrease to the stationary value. If the active region of the device is short and
comparable to the distance over which the overshoot effect takes place, electron transit
through Ihis active region will occur at a velocity higher than the stationary velocity and
the overall transit time will be shorter. Consequently. the device will be able to operate
at higher speed and frequency.
As follows from Fig. 6.13. the characteristic time of the overshoot effect is about
Ttr = 0.5 x 10- 12 S for GaAs. Estimating the average velocity under the overshoot as
Urn = (2-4) x 10 7 em S-l , we obtain an estimate for the device length necessary to realize
the discussed effect: L.t :s; UrnTlr = 0.1-0.2 ~m. That is, the overshoot effect and ultra-
high-speed electrons are achievable in short (submicron) samples.
At high biases and currents. the space-charge effects should be taken into account
together with the velocity overshoot. In Fig. 6.14, we present the calculations of space-
charge-limited electron transport in a short n+- j -n+ GaAs diode, where n+ are heavily
doped contacts and i is the undoped base of the diode. The step-like doping of the con-
tact regions is marked by N D • The field, the concentration of the injected electrons and
their average drift velocity, and the average energy are shown. The electric field and the
injected electron concentration are extremely nonuniform, which is similar to the pre-
viously analyzed case of dissipative electron transport. However, for a short diode, the
velocity overshoot in the diode base is clearly observed. The avemge energy increases
in the base, and drops only in the receiving contact because of strong scattering in
the heavily doped region. Interestingly, the drift velocity reaches a value of above 4 x
107 cm S-l and the electron energy exceeds 3500 K at a crystal temperature
of77 K.
For the low-field, hot-electron, and overshoot transport rcgimes studied, the processes
of dissipation playa dominant role in controlling the electron drift velocity. By scaling
down the sample dimensions, one can achieve an ultimately fast collision less electron
transport.
Classical ballistic transport

According to Eq. (6.11), the classical ballistic (collisionless) regime occurs in very short
samples, where during their flight the electrons do not undergo any scattering. The finite
electric current and electric resistance in such a case occur exclusively due to the space-
charge effects. We can account for these effects in a ballistic diode with the geometry
depicted in Fig. 6.2(b). Similarly to Ihe case of the above-presented short diodes with
dissipative and overshoot transport, the distribution of the electrostatic potential, <1>(x).
the concentration, n(x}, and the velocity, v(x}, have to be determi ned under collisionless
electron flight. We will exploit Poisson's equation (6.68) and usc the current density in
the form
J = -e/l(x)v(X). (6.8Z)
The velocity, v, can be found from the Newton equation (2.8); however, we will use the
cnergy conservation law, which reads as
m·v2 (x} ml);
Z - e<P(x) = 2 - e<P, = constant. (6.83)
Here, I)c and <1>c arc parameters corresponding to thc injecting electrode (cathode). In
the case ofa large electric bias, we can simplify the latter equation by considering that
the electrons are injected over the barrier with small velocity I)c --jo O. Then, we exploit
the virtual-cathode approximation discussed previously_ It is convenient to set <be = 0
and obtain
u(x) = jze<p(x)
11/.
(6.84)
and
o J
I/(x) = eu(x) (6.85)
where Jo is the absolute value ofthe current density through the diode. As we noted in the
section devoted to dissipative transport, J is negative and it is convenient to introduce an
absolutc value of currcnt dcnsity, Jo, which is constant throughout the entire sample. On
combining thc lattcr formula with Poisson's cquation (Eq. (6.68)) we obtain the equation
for <P(x):
Jo ",*
(6.86)
<0< Ze<P(xY
On multiplying both sides of the last equation by d<b(x}/d\', we obtain an equation that
can be integrated. The result of integration is
~(d<P)2 = Vo ~ + C. (6.87)
2 dx €o€ V2;-
The integration constant, C, can be set to zero, since within our approach as x ~ 0 we
have d<P/dr -+ 0 and <t> -+ O. Now, we can rewrite the result as
I d<l>
-
JO [;;,,0
- (6.88)
""<1>7'/"-' -d-, = 2 l"oE' 2e .
Further integration gives us the final result for the distribution of the electrostatic
potential:
3 ~
et>(x) = _ (J. )'/]( ~0)'/] x 4/ 3 •
(6.89)
2 fOE' 2e
Taking into account thai ¢(L.r ) = <1>0. we can find the current-voltage characteristic of
(2)]/'. g
the ballistic diode:
J b = fOf 2e <1>3/2. (6.90)

o. 3 L2x /1/. 0
This equation is known as Child:~ law. It was found and verified first for vacuum diodes,
where collisions of electrons were totally absent. Child's law differs considerably from
the current-voltage characteristic orthe dissipative diode given by Eq. (6.73).
Using Eq. (6.89), we can calculate other parameters ofthe ballistic diode. For example,
the transit time for the electrons is
~
1II0
lOb
•
= L .or - -.
ecl>o (6.91)
with IO.b being the transit time of the ballistic electrons in the absence of space·charge
effects. The space-charge effects increase the transit time, (lr.b, by the factor 3/2.
It is instructive to compare the results obtained for space-charge-limited transport in
the dissipative and ballistic diodes. To make this comparison we assume that the lengths
of the diodes, L.t , are equal and that the two diodes are biased equally. 111en. from
Eqs. (6.75) and (6.91), for the ratio of the transit times we obtain
Itr.b 9 Jm'"tpoJ-t 9T e Jecl>o 9/e
- = = = - « I. (6.92)
Itr.d 4J2iL.t 4J2",tL;c SL;c
where we used the relationship J-t = eTe/III'", and introduced the maximum mean-free
path, 'e: = TelJm , and the maximum electron velocity in the diode, Urn = J2e<J>o/m t , The
ratio obtained is small, in accord with the condition of dissipative transport of Eq. (6.12).
That is, the ballistic diodes can provide much faster regimes of operation. Similarly, it
can be shown that, for a given electric bias, the currents in the ballistic diode are much
greater than those of the dissipative diode: .JO.b/JO,d = ~(~)]/2L ... /Il:» 1. Hence, the
ultimate collisionless transport has a number of advantages over the dissipative transport
due to the higher velocity of electrons.
In fact, in a real situation some collisions always occur. Then. this "intermediate"
case can be treated numerically by using a much more complex model. In Figs. 6.15(a)
and 6.15(b) the results of such numerical modeling for a GaAs diode of length L... =
0.4 ~m at temperature T = 300 K arc presented. A typical value of mobility for high-
quality GaAs material is J.L = 7500 cm 2 V-I s-l. For this mobility, the time of free flight
(0) (b)
Contact t--Lighlly-dopcd slab- Contact
0.2 o~OV _ 14 0.8 r-~--~--~--~,'" -=."'3"'"""'"
<Po = 0.47 V x-~,I 11m
0.1 • .... .... 12
\ Co I .
\_- 10 In ,.-.. 0.6 x - 0.05 11m' ~ , .

0.0 f~----':~
~ -0.1 ~o=0.47V 8 "
E " I'
I \. .
x-!\J;!m I ~
I \
'--
.~ ~
~
• 6 -
0
. .
~
~
-0.2
... •
... •
\
•
~
~
'.
·•·/ ,I ....•.: "'

\
•
~ -0.3 •
4 ,,,
~
0.2 ... I "'. \ '
\
• /j.(x) = -1/(ell(A')} \ ''I'- .... I /' ~._':,.'~-I'~_././

-0.4 I .
2
.. .. y
.,~
. . \.
..
' ",'
0
-0.5 0 .1 0.0 0.1 0.2 0.3 0.4 0.5 0.0 •..._; 0 5 ' . 10
15
X(~Ul1) v(I01cms-~
Figure 6.15 The results of numerical modeling of a GaAs diode. (a) The distribution of electron
potential energy, e4>(x). and avemge velocity, vex). across the diode. (b) The distribution
function. F(u. x). at x = D. 0.05. D.1, and 0.3 11m. The applied voltage, 4>0, is equal to D.47 V;
the thermal velocity, UT, is equal to 2.6 x 107 em S-I. Reprinted with permission from H. U.
Bamnger and J. W. Wilkins. "Ballistic electrons in an inhomogeneous suhmicron structure:
thermal and contact effects;' P"y.~. Rev. B. 30, 7349 (1984). (0 1984 by thc American Physical
Society.
is Te = 2.9 X 10- 13 s. At the given temperature, the thermal velocity of the electrons
is ur = .jkBTj",.. ~ 2.6 x 10 7 cms-t. The average mean-free path of an electron is
Ie = 0.075 ~m and L.t / Ie ~ 5. Thus, on average the electrons with velocity ur would
undergo as many as five collisions during the flight across a uniform sample of the chosen
length, Lx. For such a sample, the transport would be almost dissipative. For a strongly
biased diode, we reach quite different conclusions. In Fig. 6.15(a), the potential-energy
profile and the average electron velocity are presented for an applied voltage <1>0 = 0.47 V.
For comparison, the thermal velocity Ur is also indicated. One can see that the maxi-
mum average velocity is above 7 x 10 7 cm S-I. The actual mean free path is estimated
to be larger, 0.07 ~m ~ Ie ~ 0.2 ~m. Thus, this example corresponds to an interme-
diate case between pure ballistic and dissipative transport. For such a case, different
electrons have different velocities and it is possible to characterize these electrons by
a distribution function, F, over the velocity, u. The distribution depends also on the
distance along the diode, x: F = F(u. x). In Fig. 6.15(b), the distribution function is
presented for various distances,x. The thennal distribution of the electrons in the cathode
is shown by the dotted line, and it is obviously the Maxwellian distribution ofEq. (6.16)
with the temperature T = 300 K. Actually, this distribution determines the velocities
of the injected electrons. The initial "spreading" of the electrons over velocities in the
cathode is one of Ihe main differences from the simple model analyzed previously, in
which the electrons are injected with a near-zero velocity. Inside the diode, the distribu-
tion becomes strongly anisotropic: electrons with negative velocities are almost absent.
As the distance increases. the distribution becomes more and more anisotropic, with a
1.0,"",,~~~-.---~~~-.---~-,
0.8
0.6
0.4
2( W1 tr,b) 0.2
2(0) O.Or------.::""-~~---i
-0.2
-0.4
-0.6h---"-*2;:::::;~"""7-"~-:'-~,
012345678910
COftr,b
FIgure 6.16 Dependences of the real (curve 1) and imaginary (curve 2) parts of the impedance of
Eqs. (6 93) and (6.94) on frequency w; Z(O) ~ Z(w ~ 0).
well-pronounced maximum. For larger distances, Ibis maximum corresponds approxi-

mately to values given by the previously discllssed model of pure ballistic transport. This
example highlights the real nature of an almost ballistic diode: some of the electrons par-
ticipate in transport and there occurs some spreading over the velocity. However, the aver-
age properties of the diode arc very close to those obtained in the simple model based on
the classical Newton and Poisson equations.
The previously discussed simple model allows one 10 investigate high-frequency prop-
erties of the ballistic diode and calculate its impedance. The real and imaginary parts of
the impedance are
12Rd
Re[Z(Il)J = 11' [2(I-cosll)-llsinn]. (6.93)
12Rd 3
Im[Z(Il)J = 11' [2sinll-Il(1 +cosll)-Il /6J. (6.94)
Here, Rd = dcl>o/dJo is the differential resistance calculated with the lise of the steady-
state current-voltage characteristic ofEq. (6.90); Q = W/tt.b, i.e., the frequency measured
in units of the inverse transient time defined by Eq. (6.91). BOIh dependences, ne[Z(Q)]
and Im[Z(Q)), arc presented in Fig. 6.16. According to Eqs. (6.93) and (6.94) and the
definition of Rd. the magnitudes of'Re[Z(Q)] and Zm[Z(Q)] decrease as the voltage
bias, <1'>0, increases:
d<l>o I 3m* L 2
Rd= - =- ---'-
dJo 2 e¢lo fOf
That is, the current response becomes larger at larger biases (Rd ex 11 J(i>;). Simultane-
ously, the frequency region within which the diode is still active increases with the bias
(0( J{i>O). These features are quite similar to those obtained for the dissipative Mott-
Gurney diode; see Eq. (6.77). A principally new feature is the observable oscillations
of'R.e[Z(w)] and 7m [Z(w)) with w and the "frequency windows" arising with negative
(a)
Lead'r-+-... Device r-+-...Lead
L R
(b) e<l>(x)
e (x) ____
t x
e<l>o/2 e<l>o/2
Rgure 6.17 (a) The structure schematic of a nanoslruclure mesascopic device and (b) the
potential energy profile, e¢(x), in the biased mesascopic device.
values of R.e[Z(w)]. The first "window" with 'R.e[Z(w)] < 0 occurs for w between
6.3jt1r.b and 9jltr.b' It is known that a negative 'Re[Z(w)] corresponds to an electrical
instability at frequency w. Such an instability can be used for generation of high-
frequency electromagnetic oscillations, and we will discuss this phenomena in detail in
Section 8.2.
In conclusion, ballistic nanoscalc devices are the fastest devices based 011 classical
electron transport. Both a finite electric current and a finite electric resistance in the
absence of electron scattering arise because of the electrostatic effects induced by redis-
tribution of charged electrons in the device. This is the reason why this case is associated
with cJwl'ge-/imited trallsport. The ballistic diodes are characterized also by "frequency
windows" within which the dynamic resistance of the device is negative, which signifies
an electrical instability and the possibility of gcncrating ultrn-high-frcqucncy electro-
magnctic radiation.
Quantum ballistic transport: the Landauer formula

The general theoretical description of the various quantum transport regimes is too
complex a problem to be prcsented in this book. We consider the simplest limiting case of
time-independent and low-temperaturc transport, whcn inelastic processcs arc ncgligible.
Thus, according to the classification given in Table 6. ( wc consider now the mesoscopic
regime of electron transport. As has been pointed out previously, transport through
nanostructure mesoscopic devices depends on both the geometry of the nanostructure
and the "'cads" (electrodes, contacts, wires, interconnects, etc.) which connect the device
to an external electric circuit. Hence, we should consider the whole system: the device
and the leads. Figure 6.17 illustrates the essential features of the simplified model. The
system consists of two leads, to which a bias is applied two "electron waveguides;' L
and R (they can be thought of as quantum wires), and the device itself.
In order to avoid the need for a detailed description of the leads, we assume that the
leads are reservoirs of electrons, where energy and momentum relaxation processes are
so effective that the electrons remain in equilibrium even under a given applied voltage
bias. Hence, the boundary conditions at the interface between the leads and the device
arc assumed to be determined by the equilibrium Fermi distribution function ofEq. (6.18).
111e electron concentration in the leads is so high that the electrostatic potential in each
lead is taken to be constant, as for the case of a metal. Let E F be the Fermi energy of the
electrons in the leads in the absence of a bias. On applying a voltage bias, 4>0, the Fermi
level in one of the leads becomes £F - te4>o. while that of the second lead becomes
£F + ~e¢o, as depicted in Fig. 6.17(b). Thus, the electron distribution functions are
h( E + ~e<l>o EF)- and FF( E - ~e<l>o - EF). (6.95)
at the left and right leads, respectively. Here E is the kinetic energy of the electrons.
In studying quantum wires in Section 3.3, we found that the electron spectrum consists
of a series of one-dimensional subbands with the energies
tl1k 1
E}".n.m -- E:
n.m
+ 2m- x (6.96)
Here, 11 and m are integers and k.r is a one-dimcnsional wavevcctor directed along the
axis of the quantum wire. The electron waveflUlction can be written as
1/I(x. y. z) = 1/1 ~(y. z) x 1/I1I(x). (6.97)
The component of the wavefunction l/! l.(Y. z) describes the transverse distribution of
the electrons in the quantum wire, while "'1I(x) describes the translational motion of
electrons along the connccting quantum wires. It is gencrally composed of plane waves
e±ak,.r. Since electron transport occurs along the wires, we must analyze the wavefunc-
tion 1/I1I(x), We can write the wavefunctiol1 of the electrons, 1/I1I(x), at the interfaces
between the leads and the device. These interfaces are indicated in Fig. 6.17 by the L
and R cross-sections. The wavefunction, 1/I1I,I(X). of the electrons coming from the left
lead is
X "'Xl,
(6.98)
x"'x p
Here, Xl and X r denote the coordinates of the cross·sections Land R, respectively. That
is, al the cross·scction L the wavefunction consists of incident and reflected waves,
while at the cross-scction R there is only the wave that has passed through the devicc.
The coefficients f r and "1 are the amplitudes of these transmitted and reflected waves,
respectively. These coefficients depend on the particular pOfenfial profile in the device.
its geometry, etc. In a similar manner. one can write the wave function, 1/I1I,r(X), of the
electrons coming from the right lead:
fle-iJl)(·r-.r l ). x ..... XI,
1/I1IAx) =
1
e-ik,(x-x" + /'.eik,(.r-x,).
, .,v ...... X r ·
(6.99)
In Eqs. (6.98) and (6.99), kl and k r arc the wavevectors in the left and right cross-sections,
respectively.
The complex coefficients I r • rl. II> and 1', are connected by several fundamental rela-
tionships, which do not depend on the specific design of the device. The requirement of
current continuity for both wave functions, "'".I(X) and "'flAx), gives
and (6.100)
Other important relationships arc
and (6.101)
Then, using Eq. (6.100), we gct
(6.102)
By substituting the w3vefunctions of Eqs. (6.9N) and (6.99) into Eq. (3.10), which
defines the particle flux, we calculate the incoming, ;In. and outgoing, ;out, electron flows:
(6.103)
where lJI = t,k1/m- and Vr = "kr/m- are the velocities at the Land R cross-sections,
respectively. The ratio of these quantities defines the ttrJllsm;SS;01J coefficienT for
electrons moving through the device from the left to tbe rigbt:
i out kr 2
1i_,(E) = -. = -k 11,1 . (6.104)
'm 'I
The transmission coefficient corresponding to the elcctrons moving tbrough the device
from the right to the left equals
k,
T,_,(E) = k,IIII'. (6.105)
From Eqs. (6.102), (6.104), and (6.105), one can find thaI
(6.106)
wbere
h'k'
E" - ----'- (6.107)
2m-
is the kinelic energy corresponding 10 the longitudinal component of the electron"s
momentum. Thus, the transmission coefficients are the same for both directions of
incoming electrons. The ratio of reflected and incoming electron fluxes defines the
reflection coefficient:
R(E)
'r = Irll-, = Ir,I-·,
= :- (6.108)
1 m
It is obvious that
T(E) + R(E) = I. (6.109)
ow, we can take into account contributions to tbe total current from all electrons
entering tbe device from both leads. Considcr a state of tbe e1cC1rons, say in the left
lead, with quantum numbers k." n, and m. The number of electrons in tbis state is given
by 2FF(E(k.r . II, m) + 1e<J>0 - E F). where the factor 2 comes from spin degeneracy. If
the connection length is L e , the number of electrons per unit length of the connection is
2Fd L e • The lotal contribution to the electric current from the electrons entering from
the left is
I, = - 2e L L"II T(E I ).1",(E(ks. m) + ~e"'o - E,).II. (6.110)

Le ".", *,>0 2
Similarly, for the electrons from the right lead one gets
I, = 2e L L ""T(EI).1",( E(k s • 11.11I) - ~e"'o - E,). (6.111)

Le ".",*.<0 2
Therefore, the total current through the device is
I = II - I r
= - ~: L l:>II T(E )~,( E(k,. 11.11I) + ~e"'o - E,)

".", *, >0
-.1",(E(ks • II. 11I)- ~e"'o-E')]. (6.112)
Since, for our model, the electron velocity, vn, and the transmission coefficient, T, are
independent of the transverse quantum numbers, II and m, we can calculate the sum over
1/ and m. Taking into account the explicit form of the Fermi function, Ff' it is convenient
to introduce lhe distribution function dependent only on the kinetic energy Ell:
I
.1",,(EI )=2L (E e -E.)' (6.113)
II.no I + exp II + k:'; F
Next, as in Eq. (6.30) the summation over kx in Eq. (6.112) may be replaced by an
inlcgmtion:
'"L..,!--
*
·1- f L,
2rr
dk,( ... ) = L,
- f dEli (... ).
2rr twU
(6.114)
•
Finally, we oblain the following expression for the total currenl:
f
1= -e ~:~ T(E I )[.1"'(Ell + ~e"'o - E,) - .1",(EII- ~e"'o - E,)].
(6.115)
where the integration range runs over the kinetic energy ofthc longitudinal motion En.
ote that Ilteeleclron ,'elocilydoes 1101 appear illlltefillal e:rpressioll for the total current.
I. This general result may be applied to a variety of different cases. Let us consider two
of them.
Adevice macroscopically large in the transverse directions

In this case the transverse quantum numbers arc the wavevectors n = k,. and m = k=,
and
En.",
t,1 (.1
= 2m. k,. + k;'l .
The function ofEq. (6.113) can be calculated explicitly,
FF(E II ) = S ",'kaT
, In [ 1 + exp (E -
rrh- k.T
F Ell)] . (6.116)
where S is the cross-sectional area. The term .'FF(EU) has the meaning of the number of
electrons with energy Ell' Then the total current density, J = 1/ S, is
EF - Ell + -~e<l>o)
J = -e
m·k B, T f dEli
-T(E II ) In
l+exp
( k.T
(6.117)
rr f,- 2rr fl
1 + exp
EF - Ell - -~e<l>o)
( k.T
This is indeed a useful result since Eq. (6.117) allows one to calcuJate the current-voltage
characteristic of a nanostructure device and its dependence on the electron concentration,
temperature, etc. Interestingly, we found ajillite current in this biased device, where 110
scattering occurs. That is, the device possesses an electric resistance. The resistance is
explained by two factors: (I) quantum-mechanical reflection of electron waves inside
the device and (2) the finite number of electrons which can be injected into the device.
Note that no electrical charge redistribution was considered, i.e., relatively small currents
were supposed.
Device conductance at low temperatures. The Landauer tormula

Let us tum to the gencral result ofEq. (6.112). It can be simplificd significantly for near~
equilibrium transport at low temperatures. We know that in the limit of zero temperature
the Fermi distribution function, .'FF, becomes a step function:
lim FF(E - EF) = G(E F - E). (6.118)

T-O
If the applied voltage, <1>0, is small, the difference between the distribution functions in
the sum ofEq. (6.112) is equal to
FF(E(k<- 1/. "') + ~ e<l>o - E F) - FF (E(kx- 11. "') - ~ e<l>o - EF) = e<l>08(E F - E).
(6.119)
where we have taken into account the fact that the derivative of the step-like Fermi
distribution function is a 8·function. Hence, the current is proportional to the voltage
bias, <1>0.
One can introduce the canductance ofa nanostructure device as the ratio
I
G--
- <1>, (6.120)
From Eqs. (6.112), (6.114), (6.119), and (6.120) we obtain the conductance at lowtem·
peraturcs in the form
el e2
G = - L T(EF,II. m) = 2- LT(EF.n.m). (6.121)
h I/.m.s h lI.m
where the sum extends only over the electron states (fI, m) with energy E < EF. The
coefficient in front of the sum of Eq. (6.121),
e'
Go =-. (6.122)
"
is called the quaI/film of com/llc/(Ulce; here It is Planck's constant; II = 2rr ft, where tl
is the reduced Planck constant introduced previously in Chapter 2. The quantum of
conductance is equal to Go = 39.6 IlS (I siemens::: IS::;; 1 A V-I) and its inverse va llie
is I ICo::: 25.2 kQ. Equation (6.121) is often called the LOlldouerjormllla.
Sometimes it is convenient to consider electron states corresponding to different quan-
tum numbers (II. m) in terms of separate electron conduction channels. In this la"cr
formulation, Eq. (6.121), may be rewritten in the form
G = 2G o L: T".m with T".m = T(Ef, II, III). (6.123)

n.m
Here each channel (11. m) contributes GoTn . m to the conductance, G. If the channel cor-
responding to the (n. 111) state is transparent to electrons, Tn.", = I, and the contribution
of this chauncl is equal to the quantum of conductance of Eq. (6.122).
This regime is also referred 10 as qllalltlllll ballistics. Thus, a nanoscale device with
quantum ballistic transport exhibitsjinite conductallce (finite resistallce). Generally, the
conductance depends both on the transmission coefficient of the device and on the Fermi
distribution functions in the leads. In particular, even if the device is entirely trans-
parent, T = I, the conductance remains finite and is equal to the quantum of conduc-
tance, Go. An increase in the occupation of the upper low-dimensional subbands gives
rise to an unusual behavior of the conductance. Indeed, at low temperature, actually at
k a T « E,l, only those subbands for which the energy of the subband bottoms En •m < £,1
arc occupied. If, on changing the Fermi energy, EF. a new subband starts to populate,
tlle conductance of the device increases in a step-like manner, indcpendently of how
many electrons occupy this new subband.ln conclusion, the Landauer formula describes
quantwn transport in mcsoscopic devices. It is valid at low temperanlre and small voltage
bias.
The simplest device with quantum ballistic transport is the so-called qllllllf//m point
contact. It can be fabricated in a number of fashions. The essence of the quantum point
contact is that two electron reservoirs are connected through a conducting region with
transverse dimensions, L= and L y , comparable to the de Broglie wavelength of the
electrons. In Fig. 6.18(a), a sketch ofa point contact on the (x.y}-plane is depicted.
The electron transport and current occur along the x direction. For simplicity, we can
assume that the channel has a constant dimension, L.H in the x direction. If the variation
of the channel dimension in the y direction, L)'(x), is adiabatically smooth, we can use
Eq. (3.50) for the one-dimensional energy subbands:
t,2]f2 n 2 t/ 2 ]fl",2 1l 2k 2
'L'
E'.. ,.m = 2m: + 2m'[L )' (x )J' + 2m·x (6.124)
where we used approximate quantization of electron motion in the y direction. The

resulting potential profiles are shown in Fig. 6.18(b), for a few subbands in the structure
(a)
y = O.5d+ ax 2
---~
-
y - -(O.5d + ax')
(b) 11/= 1,2,3 E
Agure 6.18 Geometrical representations of point contact (a) and quantized energy £1 ... in the
region afpoin! conlact (L: « L,_,n"n) (b). It is shown schematically lhat electrons can be in all
three subbands in the region with the densest hatching. Electrons can be only in the subband £1.1
in the region shown by the lcast-dcnsc hatching.
with L= « L)',nlll1' For the indicated Fermi energy, we have only one channel open for
electron motion and the second channel has a small barrier. As soon as the Fermi energy
reaches the bottom of the next subband, a new channel opens for conducting electrons
and the conductance of this point contact undergoes a step-like increase. Typically.
such narrow constrictions clearly demonstrate the quantization of conductance. Since
the potential barriers in the constriction are quite wide. tunneling processes have small
probability. As a result, each of the channels is either almost open (the transmission
coefficient T = I) oralmOSI closed (T = 0); i.e., one will expect sleps in the conductance
close to their universal value 2G o.
Consider as an example the narrow constriction defined by shallow etching on the
AIGaAs/GaAs struchlre with two·dimensional electrons. In order to perform the shallow
etching with well-controlled edges. a 13-nm-thick Al mask was formed by electron-
beam lithography and lift-off, using a field-emission scanning electron microscope at
an acceleration voltage of 2.5 kV The sample was then shallow etched, as described in
Chaptcr 4. After the etching mask was removed, an AI gate electrode was deposited,
covering the constrict'ion. By applying a voltage to the gate. one can controllhe electron
density and, thUS, the Fermi energy. A scanning electron micrograph of a shallow etched
constriction is shown in lhe inset of Fig. 6.19. The shape of the etched constriction
is characterized by two parameters: (I) the width of the constriction, d. and (2) the
curvature. a, of the parabolas describing the shallow etched walls of the constriction
16
14
12
10
~
~ 8
~ 31 K
" 6f--
20 K
41-""< ..,.~ r~
II K
2e---'r--J-----------
UK
o 0.6 0.7 0.8 0.9 1.0
<1>0 (V)
Flgure6.19 The quantization of conductance. G. at temperatures T = 1.3. 11,20. and 31 K.
Inset: a shallow etched constriction. From A. Kristensen. J. B. Jensen et 01.. "Conductance
quantization above 30 K in GaAIAs shallow·ctchcd quantum point conlacts smoothly joined
10 the background 2DEG:' J. Appl. Pllys, , 83, 607 (1998). Reprinted wilh permission from
A. Kristensen. J. B. Jensen. M. ZatTalon. C. B. Sorensen. S. M. Reimann. P. E. Lindc1of,
M. MicheL and A. Forchel, J01lrnal ofApplied Physic,f, 83. 607 (1998). Cl1998 American
Institute of Physics.
seen in the picture. If the Ix. y}-coordinate system is placed in the center of symmetry
of the constriction, the etched walls are described by
,
y = ±(O.5d + ax-). (6.125)
The quantum contact presented in Fig. 6.19 has the parameters d = 50 nm and
I
(I = 0.00125 nm- . Using the second term in Eq. (6.12-1), we can estimate the energy
distances between the lowest subbands in the narrow cross-section of the constriction;
they are 20 meV, 16 meV, 11 me\'; etc. The sample exhibits well-defined conductance
quantization at low temperatures, as depicted in Fig. 6.19 for a series of measurements at
various temperatures. The rise of the conductance for each step is close to the universal
value 2Go.
Single-electron transport
In the previously discussed cases of electron transport, \VC assumed that the nunlber of
electrons participating in this transport is so large that the discrete nature of the elec-
trons does not matter. For small devices operating with weak currents, this assumption
is no longer valid. Instead, a new type of charge-dependent electron transport, sillgle-

electron transport, develops. Below we present the basic concepts of this transport
regmlc.
In general, in bulk-like materials and macroscopic devices the electron discreteness is
not manifested in average characteristics such as the local electron density. total electric
current, etc. However, it is well known that this discreteness manifests itself by con-
tributing to current noise (fluctuations) even for macroscopic samples. For example, the
so-called shol lIoise is entirely due to the discreteness of the eJeclron charge; it arises
because of the random process of electrons entering the device.
When the dimensions of the devices are scaled down, the role of the discreteness
of charge increases. In the case of ultra-small devices, the discreteness of the electron
charge gives rise to principally new effects in electron transport. This transport becomes
correlated; i.e .. a transfer of one electron through the device depends upon the transfer
of others. The correlation appears because of the Coulomb interaction of individual
electrons. A new class of devices, referred to as sillgle-electrrm devices, is based on such
processes.
Single-electron effects rely on a charging process that occurs when electrons enter
a tiny conducting sample. When the tiny conducting sample, often called the metal-
lic island, is extremely small, the electrostatic potential of the island significantly
increases even when only one electron enters it. In general, the charging energy of a
sample is
Q'
£c=- (6.126)
2e'
where Q is the charge and C is the capacitance of the sample. For a spherical island of
radius,., the capacitance, C, can be estimated as
C = 41TfoEr. (6.127)
For example, an island with radius,. = 10 nm has a capacitance of the order of lOaF =
10- 17 F, where F is the farad - the unit of capacitance, C (I aF = I atlofarad = 10- 1& F
=
and I F I C V-I). Then, with an increase in the voltage, ~<1> (~<1> = l:!.Q/C), which
is equal to e/C with the electron charge e = 1.602 x 10- 19 C, the increase in the cnergy.
~ £ = e ~ <1>, reaches 16 meV This is comparablc to the "thermal noise energy" at room
temperature, kaT ~ 26 meV. If one electron is transferred to the island, the Coulomb
repulsion prevents additional electrons from entering the island unless the island potential
is intentionally lowered by an external bias. If the island potential is lowered gradually,
the other electrons can enter the island only one by one, with negligibly small power
dissipation.
We can define a Single-electron device as a nanostructure with so small a capacitance
that a single electron added to the device generates a measured voltage change on the
device. The basic properties of single-electron devices can be described by using the
example of a customary system: an insulated tunnel junction, I, between two conducting
clcctrodes, M (heavily doped semiconductor regions or metals). as shown in Fig. 6.20
for a metal-insulator-metal (M-I-M) system. Lei this junction be characterized by the
-lOA
e•
1
Tunnel
junction
Agure 6.20 The simplest M-I-M (metal-insulator-metal) system of single electronics: a tunnel
junction with a small capacitance, C. Single-clectron transfer is shown schematically.
capacitance, C. and the conductance, G. We can suppose that the capacitance is roughly
proportional to the cross-section. S, of this junction. Thus, a small S implies a smaJl
capacitance. The conductance of the junction, G, is small enough for one to consider
the system as a leaking capacitor. Let the system have a charge, Q. Then, when a single
electron leaks through the insulator, this event changes the initial electrostatic charge
just by the elementary charge, e: Q ~ (Q - e). Hence, according to Eq. (6.126), the
electrostatic energy of the junction changes by
t1E = (Q - e)' _ Q' = _ e(2Q - e)
(6.128)
c 2C 2C 2C
The potential difference between the plates changes from
<1>, = QIC (6.129)
to
<1>, = (Q - e)/C. (6.130)
Then, the single-electron transfer leads to a voltage fluctuation across the junction equal
'0
t1<1> = <1>, - <1>, = eiC. (6.131)
and a corresponding fluctuation orthe leakage current,
t1! = G t1<1>. (6.132)
The current related to the single-electron transfer can be estimated from the following
qualitative considerations. The uncertainty relation ofEq. (3.8) between the energy and
time allows us 10 estimate the extreme limit ofthe tunneling time, Th for such a junction:
, - t1Ec "
r > -,.::,- (6.133)
In turn, the electric current

t1Q
11 = - (6.134)
t1/
Is
Ec~
Q'
{} 2C
-
-e -e12 o el2 e Q
Figure 6.21 The charging energy, Ec, for a tunnel junction with the capacitance, C. and charge,
Q, of one of the plates. The solid arrow indicates energetically favorable electron-tunneling
events. the dashed arrow indicates an energetically unfavorable event, and the dolled double
arrow indicates tunneling events without a change in energy orthe capacitor, E c . Inset: a circuit
with charging of a tunnel junction by a current source.
associated with the tunneling of a single electron (liQ = e, lit = rio and fj,Ec =
-e' /(2C» is
(6.135)
If this current, /" exceeds the current fluctuation due to the voltage f1uclliation, .6.<1',
namely
e
I, ::: tJ.I = G 6<1> = G c' (6.136)
the single-electron lmnsfer processes will control the electric current through the junc-
tion. The latter inequality leads to the criterion
-e' » G. (6.137)
"
Note that the value on the left-hand side of this criterion just coincides with the quantum
of conductance, Go = e 2 / II, introduced by the Landauer formula of Eq. (6.121).
As a result of this analysis. we can state that, if the Coulomb energy of charging of the
device is greater than the thermal energy, kB T. and the current, I, associated with the
single-electron transport is greater than fluctuations of the leakage current, then electron
transport is correlated and single-electron effects are important. The criteria for such
single-electron tnmsport can be formulated as follows:
e'
C«- and (6.138)
kBT
i.e., both the capacitance and the leakage conductance of the device should be small.
Now we can discuss the physics of single-electron transfer by using the electric circuit
shown in Fig. 6.21. We assume a situation in which the tunnel junction in the circuit is
charged by a current source, Is. Starting with no net charge on the capacitor plates at
time I = 0, the current source slowly begins to charge the junction. For a given charging
rate and short time scales, the excess charge on the capacitor plates Q < e. Importantly.
e
--
- -...
C--;;il--f---
o +..L
C
<1>0
Figure 6.22 The current-voltage characteristic under Coulomb blockade. Within the voltage-bias
range from -eiC to +elC the current is suppressed by Coulomb correlations.
macroscopic plates can be charged with a charge less than the charge of a single electron
just by shifting the conduction electrons in the plates with respect to the positive ions of
the lauice (the so-calledpolarizariol1 charge). When the current source has charged one of
the pi ales of the capacitor 10 a charge equal to +e/2 and another plate to -ell, the plates
have the charge difference of Ie between them. The electrostatic energy of the junction
reaches the value marked by the dot in Fig. 6.21. Now I e charge can tunnel through the
junction as shown by the dotted arrow in Fig. 6.21. Thus, before the tunneling the initial
charges on tJle plates were (+eI2, -eI2), whereas after the tunneling the charges are
(-eI2, +eI2). We see from Fig. 6.21 that the charge transfer does not change the energy
of the junction. The two charge configurations are, actually, identical. Because of this,
nothing can stop the electron from tunneling back to its initial state and so on. However,
the current source continues to charge the plates and the tunneling shown by the solid
arrow becomes energeticalJy favorable (1::1 Ec in Eq. (6.128) becomes negative), and it
becomes energetically unfavorable for the electron to tunnel back to its initial capacitor
plate. llllls, in reality, we obtain tunneling of a single electron and the system starts
to accumulate the charge to provide for the tunneling of the next electron, which will
happen after some timc has elapsed. The succession of elcctron·tunneling events means
that the electron flux is highly correlated. In Fig. 6.21 all discussed tunneling cvents
are shown schematically. From the considerations based on Eq. (6.128) onc can sce that
any electron transfer is prohibited for a small initial capacitor charge: -e12 < Q < el2
(I::1Ec for such Q is positive). This physical effect is called Coulomb blockade. If the
conditions of Eq. (6.138) are met, the charging energy plays the dominant role in the
system, the tunneling of an electron is energetically unfavorable, and at low temperature
tunneling is not possible at all (if it is blocked). This results in the specific current-
voltage characteristic presented in Fig. 6.22. The main feature of such a characteristic
is the total supprcssion of the currcnt in some finitc interval of external voltage biases,
-elC < <1>0 < +eIC.
Importantly, the manifestation of the correlated transport strongly depends on the
external circuit to which the single-electron hmnel junction is attached. Let the exter-
nal circuit impedance be Z(w). If the impedance is as low as IZ(w)1 «G OI for the
most important frequencies, w ...... II!:, ....... t? 1(11 C), then charge fluctuations in the circuit
are greater than the elementary charge, e, and all correlation effects are suppressed.
Q
: TSIoT :
e • •
+-
2
o I
e
2 , e
7: --'---
SET - /sET - I
Figure 6.23 Bloch oscillations, showing the dependence of junction charging on lime.
If the external circuit impedance is in the intermediate interval Gill «

Il(w)1 < C- I .
the junction exhibits Coulomb blockade in the bias mnge from -etC to +ejC, as
shown in Fig. 6.22, but outside this range there is no correlation between tunneling
events.
Finally, an interesting current regime occurs ifIZ(w)1 » C- I » G OI • when Ihe exter·
nal circuit can be considered as a source of a fixed direct current, I. This current causes
a recharging of the junction inside the range of Coulomb blockade without electron tun-
neling; it corresponds to a linear change in the charge with time: dQ/dt = I :::::: constant.
When the edge of the blockade range is reached, an electron tunnels through the junction.
The system finds itself again in the blockade range - near the opposite edge - and the
process repeats; see Fig. 6.23. Thus, one gets temporal oscillations of the charging with
a frequency, /SF:T. determined by the current:
I
fSET= -. (6.139)
e
This is the so-called/requency o/si"gle-electron funneling (or Bloch) oscillations.
The previously discussed nonlinear current-voltage characteristics, oscillations, and
other single-electron effects give rise to a principally new approach to low-energy elec-
tronics. This field of single-electron devices is developing rapidly and has many potential
applications. Devices such as the single~electron transistor, the turnstile, and the single~
electron pump, have been proposed and realized on the basis of these effects. Though
these results were achieved at low temperatures, modern technology portends theirexten-
sion to liquid-nitrogen temperature and even room temperature.
6.6 Closing remarks
In this chapter, we have shown that a number of different electron-transport regimes

can occur in semiconductors and their nanostructures. ll1Cse regimes are characterized
by different values of the electron velocity and magnitude of the current, very different
current-voltage dependences, etc. The electrons can behave as semiclassical particles,
or as quantum ones. If the device dimension along the current is much larger than the
electron de Broglie wavelength, electron motion, typically, is classical. It can occur in a

dissipat-ive manner when electrons undergo multiple collisions caused by crystal defects
and lattice vibrations. TIle rate ofthese collisions determines the average electron velocity
achievable at a given electric field.
If the field is small, the average velocity is a linear function of the field. The coeffi-
cient in the linear relationship between the avcrage velocity and the field is the electron
(hole) mobility. In perfect materials and structures the mobility is high and limited
by scattering by lattice vibrations. For such systems, at low temperatures, at which
these vibrations are reduced. the mobility reaches a maximum magnitude that is lim-
ited by scattering by impurities. The mobility is one of the most important charac-
teristics for electronic applications. The electron concentmtion, mobility, and geome-
try of a sample detennine its electrical conductance or electrical resistance. We have
analyzed modification of the transport in oscillating electric fields and found thai the
current response is defined by a complex and frequency-dependent characteristic - the
impedance.
Then, we considered the behavior of the electrons in high electric fields, explaining
hot-electron effects and transit-time effects. Among the latter, the velocity-overshoot
phenomenon, i.e., achievement of very high speed for a shoTt period of time, is the most
important for ultra-high-speed devices.
Pract"ically, the transient time effects, including velocity overshoot, can be realized in
short devices. In short devices, another type of physical effects should also be considered.
Specifically, the formation of space charge affects the current. The electron space charge
increases the electric resistance; this case is known as space-charge-limited transport.
We have considered and compared several short diodes of different lengths. from the
dissipative diode to the ballistic device. We have found that their electrical properties are
markedly different from those of bulk-like samples; in particular, the current-voltage
characteristics are essentially nonlinear. Calculations of the impedance, which deter-
mines the high-frequency properties of these diodes, showed that the range of frequen-
cies within which the device response is high is directly related to the time of electron
tmnsit through the device. Thus, ultra-high frequencies can be achieved only for very
short (nanoscale) devices.
We have analyzed the simplest types of quantum transport and have come to a very
unusual conclusion - the quantum device possesses a finite resistance (conductance)
even if scattering is entirely absent. The physical reason for such a finite resistance in the
absence of scattering is quantum-mechanical reflection of electron waves from the inter-
faces between terminals and the quantum device. We have discussed the quantization of
the conductance in a quantum device with essentially one-dimensional electron motion.
The conductance quantization is related to electron energy quantization and manifests
itsel f as a muhiplc-step-like behavior of the electric current.
Finally, we have considered electric currents in devices so small thai the electrons
can be transferred only one by one, because of Coulomb repulsion. The associated
single·electron-transporl regime has very unusual properties, which can be exploited in
ultra-small electronic devices.
6.7 Problems 215
Additional information on the various regimes of electric current can be found in
K. Hess, Ad\'Ollced Theory o/Semiconductor De,l;ces ( ew York, Wiley-J EEE Press,

1999).
D. K. Ferry, SemiconducTors (New York, Macmillan Publishing Company, 1991).
I. I. Ipatova and V. Mitin,/ntroduclion 10 Solid Slate Electronics (New York, Addison-
Wesley Publishing Company, 1996).
V. V. Mitin, V. A. Kochelap, and M. A. Stroscio, Quantum Heterostrucfllres (New
For much more infom13tion on quantum transport and its applications to device physics
the reader can consult the following references:
R. landauer, "Electrical resistance of disordered onc·dimcnsional lattices:' Phi/os.

Mag., 21, 863 (1970).
H. Grubin, D. K. Ferry, and C. Jacoboni (Eds.), The Physics ofSubmicron Semicon-
ductor Del'ices (New York, Plenwn Press, 1989).
C. Jacoboni, L. Reggiani, and D. K. Ferry (Eds.), Quo",,,m Transport in Semiconduc-
tors ( ew York, Plenwn Press, 1992).
Discussions of charge-limited transport and single-electron transport are presented,
for example, in the following references:
M. A. Lampert and P. Mark, Currentlnjectiol1;'/ Solidr (New York, Academic Press,

1970).
K. K. Likharev, "Correlated discrete transfer of single electrons in uhrasmall tunnel
junclions,"IBM 1 Res. Develop., 12, 144 (1988).
H. Koch and H. Lubbig (Eds.), Single-Electron runneling and Mesoscopic Devices
(Berlin, Springer-Verlag, 1992).
6,7 Problems
1. Consider the bulk crystal of GaAs with an electron effective mass m- = 0.067mo,
where mo is the free-electron mass, and mobility J-L =
lOS cm 2 V-I S-l at T =
77 K.
Using the relationships introduced in Sections 6.2 and 6.5, and Eq. (6.50) (/1 =
-eTe /III-), calculate
the de Broglie wavelength, ),,;
the scattering time, Te ;
the thennal electron velocity, tiT;
the mean free path, Ir ; and
the diffusion coefficient, D (D = tI} Te/a. a = 3 for bulk material).
Dctcnninc the transport regimes for devices with feature sizes Lx = 0.05, 0.5, and 5 J,lm.
2. Consider a bulk crystal, a quantwn well of thickness 10 nrn, and a quantum wire of
cross-section 10 nrn x 10 nm. For such samples, the specific electron concentrations of
10 18 cm- 3 , 10 12 cm- 2 , and 106 CI11- 1 correspond to the same three-dimensional electron
density. For these three cases, by using the results of Section 6.3, calculate and compare
values of
the Fermi wavevector, kF, and corresponding de Broglie wavelength (J... = 211" I kF):
the Fermi energy, EF; and
the Fermi velocity, tJF = J"2"'E~F"'/-::III:O·.
3. Assuming an ambient temperature T = 4 K, prove that the electron gas considered

in the previous problem is degenerate for all three cases. For a nondegenerate gas the
thermal energy is £, = (exI2)k B T, where ex = 3, 2, and I for bulk, quantum well, and
quantum wire, respectively, and thc thermal velocity is Vr = J2E rim·. Calculate these
parameters for further comparison. Using the Pauli exclusion principle, explain why the
Fermi energy and the Fermi velocity of the electrons are higher than the corresponding
characteristics Er and Vr for a nondegenerate gas of particles with the same Illass /11 •.
4. As studied in Section 4.4 for many semiconductor materials, the energy spectra of
the holes are degenerate and consist of two branches: the light holes, Elh = tI2k2/(2m~h)'
and the heavy holes, Ebh = t,2k21(2"'h.h)' Let both hole subbands be populated, with
concentrations "Ih and "hh, respectively. At low temperature, the holes in both subbands
have the same Fermi energy. Find the relationship between the concentrations of the
heavy and light holes.
5. The density ofstates as a function of the energy has a staircase character for quantum
wells. Using Eq. (6.45), estimate the height of the stairs for the density per unit area in
the case ofa GaAs quantum well.
6. Use the energy spectra oflight and heavy holes presented in Problem 4 and compare
the densities of states for the light and heavy holes in bulk samples. Which of these two
hole subbands is more populated under equilibrium conditions?
7. Consider a short diode of length Lx with space-charge-limited djssipative transport.
Let <1>0 be the voltage applied to the diode. To make estimates for this dcvice onc should
specify the criteria of validity of the concept of the dissipative diode. For this, one should
combine Eq. (6.12) and the condition of the absence of electron heating in the diode,
i.e., F(x) < Fh' with F", defined by Eq. (6.81).
Using the estimate for the field F ::: I'f>ol Lx, prove that the diode concept works for
the voltage kB Tie < <1>0 < L:{ Jk B T l(eJ.LTE). Show that these criteria can be met under
the condition Lx> JJ.LTt.kBTle.
Assuming room temperature and the material parameters 11. = 103 cm 2 V-I S-I and
TE = 10- 10 s, calculate the criticallenglh, L;f" For L .... = 5 ~.un and <1>0 = 0.1 V, determine
the electric current density, J, and the electron transit time, Ttr.
8. Consider a GaAs diode. in which the electrons have the effective mass /II. =
0.067mo. Assume liquid~njlrogen temperature, T = 77 K. For such a temperature in
high-quality GaAs material, the typical mobility is 11. = lOS cm 2 V-I S-I.
6.7 Problems 217
U(eV)
-50 -10 0 10 50 x (nm)

Agure 6.24 A potential barrier U(x) = Uo/cosh(xjd) with Vo E: 0.2 eV and d = 10 nm.
Estimate the mean free path, Ie: = Tc: X ur. Sct the diode length L~ = Ie; i.e., asswne
that the diode is a ballistic device.
Calculate the current density, J, and the transit time, ftf. for the ballistic diode at <1>0 =
0.1 V and compare your result with parameters obtained for Problem 7.
9. The Landauer formula of Eq. (6.121) includes the cncrgy-dependent transmis-

sion coefficient, T(E). It is instructive to discuss this formula by using an exam-
ple for which the exact solution of the quantum-mechanical SchrOdingcr equation is
known. One such case is the electron motion in the potential of a special dependence:
Vex) = Uo/cosh(x jd) (Fig. 6.24). One can sec that this potential barrier has height equal
10 V o and the characteristjc spatial scale d. It was found that the transmission coefficient
has the form

8",· Vod 2
for ... , > I.
".
_81:::/1-;'U",oc::d,--'
2
sinh (rrk"d) + cosh 1 }-, -
h'
- I
\
where kx = .12m· E/ft is the wavcveclor of the incident electron. Assume thai only the
one-dimensional subballd with 11 = til = 1 is populated by electrons. Set the following
parameters: Vo = 0.2 eV, d = 10 nm. and m· = 0.067mo. as for GaAs.
Using the dependence T(E), calculate the conductance, G, as a function of the Fermi
energy.
7 Electrons in traditional
low-dimensional structures
7.1 Introduction
Now, we begin our analysis of novel developments in electronics that have resulted from
the use of nanostrucnlres in modern electronic devices. Importantly. the attributes of
nanotechnology make it possible to pursuc both devices with smaller dimensional scales
and novel types of device. Though the ongoing trend of miniaturization in electronics
is extremely important. the unique properties of electrons in nanostructures give rise
to novel electrical and optical effects, and open the way to new device concepts. The
electric current and voltage in a device are determined by two major factors: the con-
centration and the transport properties of the charge carriers. In nanostructures, these
factors can be controlled ovcr wide rangcs. In this and the next chapter wc will study
nanostnlctures for which thcse basic factors that are important forthe electronics areengi-
neered, which arc being exploited intensively both in research laboratories and in practical
nanoelectronics.
To distinguish the nallostructures already having applications from the ncwly emerging
systems, we refer to the former as traditionollow-dimensiol/al srruC/IIres.
7.2 Electrons in quantum wells
In this section, we consider a few particular examples of nanostructures with l'.vo-

dimensional electrons.
As a basis for the further analysis, we will recall and develop several of the previously
introduced definitions and properties of an electron gas. In what follows, the effect of
the band offset arising at ajunction of two semiconducting materials, which was defined
in Section 4.5 via electron affinities, is critically important. As discussed previously, the
electron affinjty is the energy required 10 remove an electron from the bottom of the
conduction band of a specific material to vacuum. The difference in affinities of two
adjacent materials is equal to the conduction-band offset at their heterojunction. Another
useful parametcrofa material is the workjiUlcriol1, which is dcfined as an cnergy required
10 remove an electron from the Fermi level of the material to vacuum. Finally, we will
use also the fundamental property of electron statistics which implies that in equilibrium
the Fermi level is constant throughollt the whole system.
7.2 Electrons in quantum wells 219
(a) (b)
Vb(Z) Charge lransfer,..~
-00000- ~----
-e(z) <>01J)(J:/!!~ t-f £
mEG F
0,
£
Unstable state - Thennal equilibrium Z
Figure 7.1 The charge-transfer effect in a selectively doped single heterostructure. (a) The
step·like profile is unstable. (b) Charge transfer results in ionization of a layer of the doped
material, band bending. lining up the Fenni levels across the structure, and fomlulion of an
electron channel at the interface.
Single modulation-doped heterojunctions

First, let us analyze qualitatively a single semiconductor hctcrojunctioll. We consider a
junction of two semiconductors with affinities such that they result in a discontinuity
of the conduction band as indicated by Fig. 7.1 (a). Then, we assume that the barrier
semiconductor material is n-doped while the narrow-bandgap material on the right-hand
side of the structure is undoped. In real situations, the latter material is, usually. lightly
doped by acceptors; i.e., it is a p-type material.
The doped regions ofthe system fix the positionsoft11e Fermi level, EF. At equilibrium,
the energy levels below EF are occupied by electrons. Accordingly. the scheme sketched
in Fig. 7.1 (a) is unstable. The electrons will move toward the undoped crystal until an
electrostatic field, brought about by the redistribution ofelectrical charge, bends the band
edges so that the Fermi level becomes constant across the materials; i.e., we obtain the
same Fermi level in the n-doped region and in the undoped region. Instead of having an
energy step as in Fig. 7. I(a), one obtains the situation shown in Fig. 7. 1(b). That is, the
band edges are bent, there is ionization of impurities in some region of the doped part of
the system, and there are free electrons inside the potential wcll. This well is formed by
both (I) the bandgap discontinuity and (2) the electrostatic potential. The situation just
described is frequently referred to as/ormation ofan electltJlI gas at the iutelface.
Threc important conclusions can be derived immediately from these considerations.
(i) Though both the initial materials were insulators (at least at low temperatures), now
at the interface, near the junction, one obtains an electron channel and electron
concentration that are finite even at T = 0 K.
(ii) Charge carriers in the potential well are separated spatially from their parent impu-
rities from the barrier side of the structure. Charged impurities, which usually lead
to large scattering rates and low mobilities, serve only as sourccs of carriers, and
scattering of electrons from the potential well by the impurities in the barrier is
suppressed as a result of the spatial separation of electrons from impurities.
(iii) The bending of the energy band creates a confining potential for carriers in one
direction, say along thc z-axis. Hence there is a quantization of electrons in the z
direction and the establishment ofa two-dimensional electron gas is quite possible.
220 Traditional low-dimensional structures
To complete the discussion of this simple model, leI us note that in the real situation the
side of the heterostructure with the lower conduction-band edge is usually doped lightly
with acceptors. In this case, the right barrier of the well is higherthan half of the bandgap
of the narrow-bandgap material. Of course, the residual acceptors reduce the electron
channel mobility but their effect is weak.
Here, we considered a heterostructure with the Fermi level defined only by the modula-
tion doping. These structures are called //Ilgafed heterostrtfctures. For galed helerostrllC-
flires the physical picture is slightly different. Such gated heterostructures will be studied
subsequently.
Basic equations describing the physics of the electrons at an interface

It is important to have an idea of the basic equations describing the electrons at an inter-
face. We began by considering a step-like discontinuity of the energy band and found that
the potential welt is formcd by chargc transfer in space. Thus, the shape of the potential
is determined by the charges of all electrons on the interface and ionized impurities. On
the other hand, this many-body potential determines the motion of each electron and the
total number of ionized impurities. Thus, we face the so-called se!.f-collsisfellf problem:
the potential is defined by the concentration of electrons and ionized impurities, and it,
in turn, affects their redistribution. The simplest approach to this self-consistent problem
is to treat the electron quantization in the scheme wherein the potential is described by
a self-consistent electrostatic field. The corresponding electrostatic potential, <1>(2), can
be thought of as a function dependent on only the coordinate z perpendicular to the
interface and is a solution of Poisson's cquation:
d'(z) = - e
---:-',;-:'
dz- E"oE"
[ 2I
L Iy,,,(r)
v
F(E,,) - No(z) + NA(z) .] (7.1)
Here, e is the elementary electrical charge, E" is the dielectric constant, and £0 is the
permittivity of free space. The total charge consists of the charges of the electrons
and impurities. Let the wavefwlctions of the electrons be t/tv(r), with 11 being the set
of electron quantum numbers. Then, 1't/t,,(r) 12 represents the probability of finding the
electron of the state labeled by 11 at the point r.ln order 10 calculate the contribution of
the electrons to the space charge we introduce the energy-dependent electron distribution
function, :F(E v ), which determines the probability of the electron occupying the energy
level Ell' Thus, the electron charge density is -e L" 1't/t,,(r)12 :F(£II)' The charge of the
nonunifonnly distributed positive donors and negative acceptors is determined by their
densities, eNo(z), and, -eNA(z), respectively.
We assmned that the potential, (z), does not depend on the x- and y-coordinates.
Thus, t/t,,(r) can be factorized as discussed in Section 3.2:
1 •
y,,,(r) = jSe·(,·,+k")xAz). v = {j. kill. (7.2)
where S is the area of the junction and kll = {k x . ky }. The factorization (7.2) leads to
• I
y,,,(r)y,,,(r) = 1y,,,(r)1 2 = sIXj(z)1 2
. (7.3)
7.2 Electrons In quantum wells 221
which makes Eq. (7.1) dependent only on the z-coordinate. We also obtain the one-
=
dimensional SchrOdingcr equation for the direction:
h' d' )
( --2- •. '
m· uz·
+ V.(=) - e<l>(=) Xj(=) = 'IXj(=)· (7.4)
Here, the t'Olal potential energy consists of two contributions: the built-in potential of
the heterostructure, Vb (=), and the self-consistent potential, -e<l>(:). In our case, Vb (=)
corresponds to the energy-band discontinuity at the junction - see Fig. 7.I(a) - which
depends only on the :-coordinate. We set
where 8(=) is the Heaviside step-function. In Eq. (7.4). Ej are the energies of two-
dimensional subbands. Thus, the total electron energy is given by Eq. (6.17):
ftlk 2
£/kll ) = 'j + 2m·
II
Now. we can calculate the electron conccntrntion:
(7.5)
Here, we also introduce the sheet density of electrons for the level j:
"'.1(£') =~ L F(s. kll. j) = lIl'k~ T In[1 + exp (£, - 'I)]. (7.6)

S .t~
s . .t• rr ft- k T B
Actually, this result repeats Eq. (6.116). The quantity IT •• j is a function of the temperature,
T, and takes the simplest form as T -+ 0:
III·
"'.I(T = 0) = --.,(£, - '1)0(£, - 'I). (7.7)
iTu-
Equation (7.7) indicatcs that the level j is occupied if the Fermi level cxceeds the
corresponding energy of quantization of transverse electron propagation, E j'
It is necessary to formulate boundary conditions for the coupled differential equations
represented by Eqs. (7.4) and (7.1). For localized electron states, we may set Xj(z) -+ 0
for z -+ ±oo. For the electrostatic potential we suppose that dcl>/d= -+ 0 for --+ =
±oo.
In conclusion, Eqs. (7.1), (7.-1), and (7.5) together with thc boundary conditions com-
pletc the fomlUlation of the self-consistent problem that describes the formation of a
two-dimensional clectron gas at the interface as well as the electron quantization. By
=
integrating Eq. (7.1) over an infinite range of and using the boundary conditions, one
can find a neutrality equation in the form
(7.8)
This result implies that, despite the charge transfer. the entire system remains electrically
neutral.
Vacuum level
Ca) (b)
Xl
X,
-0 O<lo-O~•
•
•
llE,
£ ,I •
•
•
: GaAs
•
£"
• •
AlGaAs •• •• £g2
•
•
•
:0-0 10-0
·1l1E,
Figure 7.2 Schematic energy band diagrams of a selectively doped AIGaAs/GaAs
heterostructure before (a) and after (b) charge transfcrs have occurred. Relativc positions of the
valence and conduction bands arc given for both materials. The electron affinities are shown
conditionally (left). The Fenni level in the AIGaAs material is supposed to be pinncd on the
donor level. The narrow-bandgap GaAs is slightly p-doped. After C. Weisbuch and B. Vinter,
QuaI/film Semicomillctor StrllCtllre.f (San Diego, Academic Press, 1991).
Numerical analysis of a single helerojunclion

Accurate analysis ofthe problem ofelectron-channel formation at a heterojunction can be
accomplished through numerical solution of the system of Eqs. (7.1 )-{7.5). Frequently,
to solve the Schrodinger equation one exploits the so-called variationlll method. Briefly,
the method is based on the usc of so·called trial It'llvefimctiolls containing a few well-
chosen parameters, which thcn arc dctermined from the condition that the total energy
of the system should be a minimum.
If we assume that only one lowest subband with j = I is occupied by the electrons, a
simple trial wavefunction can be chosen in the form
XI(=) = 0 for z :; 0 and for z ~ O. (7.9)
which implies that the wavefunction is localized near the interface and is equal to zero
on the wide-bandgap side of the heterojunction.
Before discussing the numerical results, let us return briefly to the picture of the
electron energy at the heterojunctioll in order to present both the conduction- and the
valence·band profiles for specific hetcrojunctions. The left-hand side of Fig. 7.2 shows
the band edges of separated AIGaAs and GaAs; the distance to the vacuum level is shown
provisionally. It is assumed that AlGaAs is heavily n-doped and GaAs is doped lightly by
acceptors (p.dopcd). This doping brings about the pinning of the Fermi level at the donor
level on the AIGaAs side orthe stnlcrure. On the right-hand side of the picture, one can
see the energy structure of the AlGaAs/GaAsjunetion with two depletion regions: one,
positively charged, on the AIGaAs side and another, negatively charged, on the GaAs
0.3r-~-~-: -~-~-~-,
GaAs
,
~
0.2 (i:j
> (j)
~
" (j)
~
~
(j)
(j)(j)
""
UJ 0.1
---- EF -~-
,,, ... ~_-: --

,,, ,,
0
300 200 100 0 100 200 300 400
z CA)
Figure 7.3 Calculated self-consistent potentials, energy levels, and wavcfunctiol1s of an
Al o.3 Gao.7AslGaAs selectively doped heterostructure. The junction is situated al : = O. The
spacer thickness is 50 A and the donor binding energy of AIGaAs is chosen as ED = 50 meV.
After T. Ando, "Self-consistent results for a GaAs/AI.yGul_rAs hctcrojunction. I. Subband
structure and light-scattering spectra," J. Phys. Soc. Japan,S I, 3893 (1982).
side. For such a doping the quantum well occurs only for electrons and does not occur for
holes, since the AIGaAs/GaAs hetcrojunclion is of type J (according 10 the classification
of Chapter 4.)
In Fig. 7.3, the results of numerical solution of Eqs. (7.1)-(7.5) are presented for
a selectively doped Alo.JGao.7As/GaAs heterojunction. Heavily n-doped Alo.JGao.7As
and lightly p-doped GaAs materials are separated by a thin undoped Alo.JGao.7As layer
with thickness dsp. Such an undopcd layer is called the spacer layer or simply a spacer.
In calculations, the sheet electron concentration, "s, in the charmcl at the junction is
equal to the number of ionized donors per unit area, N s = 5 X lOll cm-2 . The sheet
concentration of ionized acceptors. Nucpl, is supposed to be much smaller: Nlkpl = 5 X
10 10 cm-2 . The self-consistent potentials, energy levels, and wavefunctions arc shown
for two different cases. For the first case, shown by the dashed curve in Fig. 7.3, the
trial electron wavefunction is defined by Eqs. (7.9). The second case, shown by the solid
curve in Fig. 7.3, corresponds to a more sophisticated trial wavefunction, which can
penetrate under the barrier into the wide-bandgap part of the structure. In the region of
classically allowed motion, the wavefunctions for all ofthesc approximations are similar
to each other. However, they are drastically different in the barriers. Generally, the trial
function of Eqs. (7.9) gives a reasonable value of the energy 1:1, as well as the behavior of
the potential at large distances. However, the second trial function yields more accurate
results. In particular, it gives a slightly lower energy level. Results obtained for various
concentrations, lis. of electrons at the interface are summarized in Table 7.1.
In this table, the lowest energy level, el, the spatial scale of the electron confinement.
(z), and the probability of finding electrons in the barrier, 1\, are presented for three
Table 7.1 Results of calculations of electron parameters for

two wavefunclion approximations: I corresponds to the
function of Eqs. (7.9), and II to a more accurate trial function
"I, (1011 em-I)
2 4 8
£1 (mcY) 38.6 56.6 86.1

(z) (A) 116 99 82
Pb (%) 0 0 0
El (mcV) 32.00 45.6 66.3
•
11 (z) (A) 100 82.6 65.6
Pt>(%) 0.7 1.11 1.95
values of the electron concentration, lis- The value of (z) was calculated as the quantum-
mechanical average of the z-coordinate of an electron, and Pb = J~ dzlx,(z)12. One
can see Ihat the relative height orthe first energy level, El, increases as the concentration
increases. The value of (1) decreases with increasing lis as well as with the confining
electrostatic potential. Thus, we can see that the width of the electron channel is in the
•
range 60-100 A. The probability of finding electrons in the wide-bandgap barrier is
small, but increases with increasing electron confinement. Obviously, Pb is equal 10 zero
for the wavcfunction of Eqs. (7.9).
The examples given in Ollr discussions illustrate the following major fcanlres of a
selectively doped heterojunction: (J) the formation ofelectron·conducting channels with
concentrations in the range lis = 10 11 _10 12 cm-2 at any temperature, including T = 0 k;
(2) the spatial separation of the electrons from their parent donors with very low prob·
ability of electron penetration into the barrier - less than or about J% - as well as
spatial isolation of the electrons from the p-doped narrow-bandgap material; (3) the
formation of a potential well for electrons with the potential profile self-consistently
dependent on the electron concentration; and (4) the quantization of electrons inside the
potential well with the resultant two-dimensional character of the electron spectrum and
with the electrons confined in the two-dimcnsional channel with a width of less than
•
100 A.
Control of charge transfer

We have considered a heterojunction formed by two semi-infinite semiconductors with
fixed concentrations of donors and acceptors as illustrated in Fig. 7.2. This results in
a conducting channel at the interface with a constant electron concentration that is
determined by the doping profile. The control of the conductivity, or more exactly, the
control of the resistance - or its inverse, the conductance - of the structure is necessary
in order to realize useful devices.
Let us consider the possibility of changing the conductance of a hcterojunction
by controlling the electron concentration. For that purpose, we study the so-called
(a) (b)
Gated heterostructure
Ungated heterostructure Meta!i"'a"'tc'-_----=
n+·AIGaAs n+·AIGaAs
......................................... 2DEG .
p.GaAs p-GaAs
Agure 7.4 Schematic diagrams of (a) ungated and (b) gated heterostructures with
two-dimensional electron gas (2DEG).
(a) (b)
-£F
I I
z z
Figure 7,5 Conduction·band diagrams for M/AIGaAslGaAs heterostruetures. The built-in

Schott",. voltage connols the depiction region under the metallic gate. It results in (a) a
normally-ofT device for a narrow barrier and (b) a l1onnally-ol1 device for a wide barrier with
2DEG in the potential well (60 nm or wider).
gated hetcrojunction which is presented schematically in Fig. 7.4(b). For comparison,

Fig. 7.4(a) depicts the ungated heterojunction considered previously. The only difference
between them is the metal (M) contact placed on the top of the n+ layer of the AJGaAs
barrier material in the gated structure. This metal semiconductor system (MES) is called
a MES structure. For GaAs-like materials, MES structures are of the most importance
for device applications, because these materials do not have a stable natural oxide, unlike
in the case ofSiOz on silicon. Therefore, most of the electronic devices based on GaAs
use MES structures. One also refers to these structures as Schottky-gate stniClttres.
Typically, under a metallic gate in GaAs-like materials there arc extended depletion
regions. whjch OCCUI because of a high built-in SchotJl..y voltage, c1>b, of about 0.8 V.
Sueh a depletion region is known as a Schottky deplelion region. The conduction·band
energy diagram for an MlAIGaAslGaAs heterostructure is presented in Fig. 7.5 for two
thicknesses ofthe AlGaAs layer. The n-doped region is separated from the junction by an
undoped spacer. The structure is shown in Fig. 7.5(a) for a relatively thin AIGaAs layer,
and Fig. 7.5(b) depicts a structure with a thicker AIGaAs layer. For both cases, there
exists an extended depletion region, which affects the electron channel formed at the
AIGaAs/GaAs intcrface and provides two possibilities of controlling the structure. The
lIormally-offstructure corresponds to Fig. 7.5(a). The depletion region extends through
both a thin AIGaAs layer and the junction. The boltom of the quanhun well shifts up.
The Fermi level lies under the lowest energy subband. Thus, there are no electrons inside
the channel and the conductivity along the heterostructure is almost zero. The donors
in the AIGaAs doped region are ionized, and the electrons have left the semiconductor
part of the structure which is charged positively. In order to turn on the conductivity
of the device, it is necessary to apply a positive I'o/Iage to the metal gale. normally-off
structures can be fabricated by using a Ihin AIGaAs barrier.
The 1Iormally-oll structure is illustrated in Fig. 7.5(b). In this case, the built-in voltage
drops across a thick AIGaAs layer so that the Fermi level lies above the lowcst subband
and electrons populate the channel without an external voltage bias. This channel has
a finite conductivity under normal conditions. This case can be realized for sufficiently
thick AIGaAs layers. Thus, in normally-on devices, one can control the conductance
of the channel by applying a lIegmi! le 1I0/rage to the metal. A large voltage leads to a
depopulation of the channel and can switch the device off.
The calculated electron potential energy and quantized levels both for normally-ofT
and for normally-on devices are shown in Fig. 7.6 for various gate voltages, G. The
Fermi level is taken to be at the zero energy. The values for the normally-ofT device were
•
calculated for an AIGaAs layer thickness of about 400 A; see Fig. 7.6(a). The quantum
well formed on the interface contains up to four quantized levels. A positive voltage
lowers the bottom of the conduction band ofGaAs at the interface. The bottom touches
the Fermi level at <1>G = +0.3 V and the device is turned on at a threshold voltage of
about +0.8 V when the first quantized level touches the Fermi level. It is clearly seen that
in the AIGaAs barrier layer a potential minimum occurs and tends to be lowered with
increasing gate voltage. This phenomenon can result in a negative effect: the formation
of a second channel in this layer, which will collect electrons, screen the gate voltage,
and result in loss of control of the concentration of the two-dimensional electron gas at
the interface. The normally-on device is shown in Fig. 7.6(b). It has an AlGaAs layer
•
thickness greater lhan 600 A. The device can be switched off when a negative voltage of
about -0.5 V is applied to the gate.
The results just presented were obtained by numerical calculations. Let us consider
some experimental data related to the problem of modulation-doped heterostructures
where carrier concentrations and their mobilities have been measured simultaneously.
In Fig. 7.7 the sheet electron concentration controlled by the gate voltage is shown for
AIIAIGaAs/GaAs systems fabricated for high-electron-mobility transistors (HEMTs).
The curves correspond to various spacer thicknesses lisp. One can see that the electron
concentration can be varied over one order ofmagnilude. Saturation of the sheet concen-
tration at high positive voltage is caused by transitions of electrons to the potential well
which is formed in the middle depleted barrier region, as discussed previously. Figure 7.7
(a) (b)
300 ,-----,,-------.-----,--.--, I
I
I
$G=O.3V I $a=-J.5V
I
I
200 I
,...... I
> I
"
E
'-'
I
I
I
100
~ I
\
I
I
I
I
"" I\.
-?
\ I
UJ \ \ I
\ I \ I
,~
\
0
(~ ~O.8 V (~G~-V·5 \ /'
AIGaAs GaAs AIGaAs GaAs
-100 L----'_----'-_--'-_-'----'
o 400 800 0 400 800
Distance from gate (A)
Figure 7.6 Calculated self-consistent potentials for conduction electrons in two MIAIGaAslGaAs
helerostruclures (a) corresponds 10 the normally-off device; (b) corresponds to the nonnally-on
device at room temperature. The Fermi level is at E = O. Horizontal lines indicate the bottom
ene'b'Y of the lowest four subbands. Dashed lines show donor levels. After B. Vinler. "Subbands
and charge control in a two-dimensional electron gas field-effect transistor," AppJ. Ph}'s. Leu..
44.307 (1984). Reused with permission from B. Vintef, Applied Physics Leiters. 44, 307
(1984). if' 1984 American Institute of Physics.
shows that considerable changes in the properties of the structures occur with changes
in spacer thickness. Among the various structures presented in Fig. 7.7, Ihe particular
structure with lisp = 0 is most likely to be useful for fabricating a normally+otr device.
Indeed, by applying a positive voltage to the gate one can increase the electron concen-
tration in the channel. Structures with a thick spacer are well suited for normally-on
devices; in these struchlres, positive voltage does not change the concentration, whereas
a negative voltage reduces it sharply.
The spacer is an important element of modulation-doped structures because it partially
prevents the scattering of channel electrons by the heavily doped side of the heterostruc-
ture and increases the electron mobility. On the other hand, there is one negative effect
of a spacer. Increasing the spacer thickness leads to an increase of the potential drop
on the spacer and consequently to a lowering of the electrostatic potential that confines
electrons near the interface. Hence, a thick spacer causes a decrease in the electron con-
ccntration. ll1is trade-off bctween the mobility and the carrier concentration requires an
optimization of the structurnl design for each particular device application.
xlO Il
~IO dsp =
,
N
~
-" u
• 8
1/
d~pl
= 5 )( 10 10 cm- 2
Al o.3Gno.7AS/GaAs
oA
~
" T~ 12 K
"0
-'"
~
~
6 45 A
".u.. 75 A
"u0 4
...
~ 180 A
~
~
...u'"
u
2
.:;
~
til " 0
-2 -I 0 1
Gale voltage, l\lG (V)
Figure 7.7 Measured gate-vohage dependences of the channel density of two·dimensional

electrons in Alo.3GacnAslGaAs structures at T = 12 K and various spacer thicknesses dSfl'
All samples are Si-dopcd with No = 4.6 x 10 17 cm-3• except for the sample with 180~A spacer.
which has No = 9.2 X 10 11 cm- 3. TIle thickness of the doped AloJGao.1As layer is
70o-1200A. From K. Himkawa, H. Sakaki. and J. Yoshino. "Concentration of electrons in
selectively doped GaAIAslGaAs heterojunction and its dependence on spacer-layer thickness
lind gllle electric field:' Appl. Ph)'s. Lett., 45. 253 (1984). Reprinted with permission from K.
Hiraka\va. H. Sakaki, and J. Yoshino, Applied Physics Letters, 45. 253 (1984). Cl 1984 American
Institute of Physics.
Here, we have considered systems with a single heterojunction. These systems can
be fabricated by a relatively simple technology and they have numerous applications,
but they suffer from greatly limited carrier concentration in the conduction channel. As
can be seen from Fig. 7.7, the typical surface concentrations are less than 10 12 em- 2 for
singlc-hcterojunetion devices. Higher concentrations of carriers in the conducting chan-
nel can be obtained in a double-junction system. This system is illustrated by Fig. 7.8.
where possible double-junction heterostnlctures arc presented schematically. For het-
erostnlctures of type I, quantum wells are formed both for electrons (two quantized
levels are shown) and for holes (one quantizcd level is shown), whereas for a type-II
heterostructure a quantum well arises only for the electrons.
7.3 Electrons in quantum wires
A quantum wire is a conductive structure, wherein electron transport is constrained

primarily to be along a single direction. Let this direction be along x. For the two other
directions-the y and z directions-the quantunHllcchanical confinemellt ofthe electrons
is imposed by the heterointerface potentials, or by suitable externally applied electrostatic
potentials. Such a system is also called a one-dimensional electron system. Electron
7.3 Electrons in quantum wires 229
(a) (b)
E E
Con ducli 00
baod
Cl.t: &2.t
Vb.1:
C1.( El.I:
eo,
V" ,m1 ,
Vb.h b
t'l.h
L
-- o L z - L- 0 -L z
2 2 2 2
Figure 7.8 Two types of band diagrams of double hcteroslructures: (a) type I and (b) type II.
(a) (b)
Figure 7.9 Quantum wire formed by etching (a) and split-gating (b) of the two-dimensional
electron gas.
motion along the free x djrection is characterized by a one·dimensional wavevector k.,.

The wavefunction has the form
where W"I.n:(Y' z) corresponds 10 confined transverse motion of the electron. The bound
states corresponding to transverse mol ion are enumerated by integers 111 and 112 and
usually are called one-dimensional "subbands:' In Section 3.3, we considered electron
confinement in two directions and found subband energies of the form of Eq. (3.50).
If distances between the lowest subband and excited subbands are small in comparison
with both the thermal energy, kaT, and the Fermi energy of electrons, E F, scattering of
electrons from the lowest subband into higher subbands is relatively weak. The electrons
occupy primarily the lowest subband and behave almost as one-dimensional particles.
Currently, there are several methods forthe fabrication ofquantum wires, including the
direct growth of the wires. The most obvious method is to start with a two-dimensional
structure and to impose an additional confinement to two-dimensional electron gas, as
shown schematically in Fig. 7.9. In the case of Fig. 7.9(a), the etching process is used to
realize geometrical restrictions for electron motion. In the case of Fig. 7.9(b), additional
confinement is induced by an electrostatic potential that is applied to a mctal1ic split gate
placed on the top of the heterostructure.
12 T~4.2 K
10
wire length
~ 8 (~m)
""
~
'"'6
N
6 0.2
~ 0.4
C,) 4 I
2
2 5
7
0 10
0.8 0.9 1.0 1.1 1.2
<Pg (V)
Figure 7.10 The dependence of the conductance, G, ora quantum wire versus gate voltage. <l>g,
for seven differenllcngths of wires as indicated near each curve. After L. Worschech, F.
Bcuschcr, and A. Forchel, "Quantized conductance in up 10 20 I.ml long shallow etched
GaAs/A1GaAs quantum wires:' Appl. Phys. Left.• 75. 578 (1999). Reused with pcmlission from
L. Worschech, F. Beuschcr, and A. Forchel, Applied Physics Leiters, 75, 578 (1999). © 1999
American Institute of Physics.
Electron transport in Quantum wires

The principal distinct features of electron transport in quantum wires are related to
Landauer quantization of the conductance. This quantization can be observable for per-
fect wire structures at low temperatures. As an example, we discuss measurements of
the conductance for AIGaAslGaAs quantum wires obtained by the etching method, as in
Fig. 7.9(a). The quantum wires were fabricated on modulation-doped AIGaAs/GaAs het-
erostructures grown by molecular-beam epitaxy. High-resolution electron-beam lithog-
raphy was used to define the masks of the quantum wires. With the help of wet chemical
etching, the doped layer and the spacer- both fabricated with AIGaAs - as well as 70 nm
ofGaAs were removed. From scanning electron microscopy, the geometrical width of
the quantum wires was determined to be 135 nm. The micrographs indicated a very small
sidewall roughness of the wires. Before etching, the electrons at the selectively doped
AIGaAs/GaAs heterointeface were characterized by a density of about 3 x lOll cm- 2
and high mobility in the range (1-2)x 106 cm:! V-IS-I. At low temperature (4 K), the
electron gas is degenerate and the mean free path of electrons having the Fermi energy is
estimated to be in the range 10-20 J..un. Thus, for wires fabricated with lengths less than
10 J..lm, we can expect Landauer quantization. To control one-dimensional subbands in
the wires, an aluminum lOp gate was evaporated onto the top of the structure. A voltage
applied to this gate changed the confining potential and subband energies. The result's
are presented in Fig. 7.10 for various wire lengths. The conductance quantization is seen
clearly for the wires shorter than 5 J..l1l1. Up to ten "steps" in the conductance are seen;
i.e., about ten one-dimensional subbands can be observed by changing the gate voltage.
Experiments with magnetic field and temperature variation also reveal the quantization
7.4 Electrons in quantum dots 231
effect and allow one to determine the concentration of one-dimensional electrons. The
typical concentration and subband spacing were found to be 1110 ~ 6 X 106 cm- l and
.6.£1.2 ~ 12.5 meY, respectively. The relatively large subband spacing - equivalenl to
::::: 140 K - implies that the one-dimensional character of electron motion should be
unchanged even at temperatures of several tens of degrees Kelvin.
Forhighertemperatures, at which electron motion becomes semiclassical, Ihe transport
properties of quantum wires based on semiconductor heteroslnlctures are similar to those
of quantum wells and pure bulk-like materials. Thus, selectively doped quantum wires
should exhibit high electron mobilities under low electric fields and high drift velocities
characteristic of hot electrons under high fields.
The situation with electron transport in carbon nanotubes is quite different. In high
fields the electron drift velocities can reach magnitudes in thc range (2-4) x 10 7 cm S-l,
which are larger than those in perfect III-V semiconductor compounds. However, the
electrical properties are affected significantly by the surrounding environment, which
influences the removal of heat from current-carrying nanotubes. For cxmnple, suspended
nanotubes display drastically differenl electron transport from that of those on substrates.
Figure 7.11 illustrates this difference. In the upper panel of this figure, scanning elec-
tron microscope images of two single-walled carbon nanotubes with Pt contacts are
shown. The left nanotube segment is not sllspended and is in contact with a nitride-based
substrate. The right nanotube segment is suspended over a 0.5-l..lm-deep trench. The
diameters of the nanotubes were measured by Ihe atomic-force microscope to be in the
range 2-3 nm. In the lower panel, the results of measurements of the current-voltage
characteristics for the 3-l1m length of suspended and non-suspended segments of the
nanotube are given. The measurements were conducted at room temperature in vacuum.
One can see that the non-suspended nanotube portion displays a monotonic increase
in the current, approaching 20 I1A under increasing voltage, V, while the current in
the suspended tube reaches a peak of ~5 \..lA followed by a pronounced current drop.
This strikingly different behavior is due to significant self-heating effects of the wires
carrying ultra-high current densities. Indeed, in the case oftbe suspended tube the Joule
heating can not be removed effectively, because thermal fluxes are possible only through
nanotube contacts. As a result, the temperature of the tube increases, especially in its
central region.
7.4 Electrons in quantum dots
A quantum dot can be defined as a material system in which electrons are confined in
all three directions. Some particular examples of sucb systems were described in previ-
ous chapters. They include the self-forming nanosized semiconductor islands, clusters,
and nanocrystals studied in Chapter 5, etc. Quantum dots can be fabricated by various
methods, including a dircct way, starting from a structure with "va-dimensional elec-
trons and using an clching process to impose geometrical restriction on the electrons
in two additional dircctions. A small mctallic gate fabricated upon a two-dimensional
heterostructure also can provide the necessary electron confinement under a negative
<a)
(b) 16
14
12
10
~
« On Substrate
-
a 8
6 Suspended
4
2
0
0 0.2 0.4 0.6 08 1 1.2
v (V)
Flgure7.11 (a) Carbon nanolubcs on a substmlc and suspended over n trench; (b) their
current-voltage characteristics. Reprinted with permission from E. Pop. D. Mann et 01.,
"Negative differential conductance and hot phonons in suspended nanolubc molecular wires:'
PlJys. &1\'. Leu.. 95. 155-50S (2005). ~ 2005 by the American Physical Society.
voltage applied to Ihe gate. In Fig. 7.12, three types of quantum dots are shown: a
nanosizcd island self-forming under a special growth regime, colloidal nanocrystals. and
a "quantum box" fabricated in a controlled way from a two-dimensional heterostructure.
An example of a quantum dot formed by using the technique of fabricating a gate upon
a two-dimensional heterostructure will be discussed in the next chapter: see Section 8.4.
The most notable feature of quantum dots is that all existing degrees of freedom of
electron propagation are quantizecl if a confining potential is deep and the dimensions
ofthc structure arc comparable to the de Broglie wavelength of the elcctron, as defined
in Eq. (6.5). According to the analysis given in Section 6.2, such a system can be called
a :ero-dimensional system. The latter terminology stresses the dramatic changes of the
electronic properties in quantum dots. Indeed, in a crystalline solid as was discussed in
Chapter 4. electrons possess continuous energy bands with the occupation, width, and
(a) As,
aAs
(b) CdS
••• -
Molten Silicate
'-;":"1-=
, 8 nm
(e) AIGaAs
GaAs ..........
AIGaAs-
Figure 7.12 Schematic representations of three different approaches for the fabrication of
quantum dots: (a) self-organized growth of nanosized islands. (b) nanocrystals in a colloid, and
(c) artificial patterning and etching of a heterostructure with two-dimensional electron gas.
separation of these bands determining the fundamental electrical and optical properties
of the solid. At the other end of tile length scale, for individual atoms the electronic state
density is discrete, resulting in, for example, the absence of simple electron transport
and intrinsically sharp spectral optical lines. In some respects, the electronic structure
of quantum dots might be said to fall somewhere between these two extremes. Thus.
quantum-dot structures are like large artificial atoms, "macroatollls."
For the analysis of tile electronic properties ora quantum dot, a model ofa confining
potential may be used. In Section 3.3, we considered two examples or these models:
the quantum-box model given by Eq. (3.51) and the spherical-dot model of Eq. (3.53).
The results obtained with these models for the energy spectra and wavefunctions can be
applied to semiconductor quantum dots, ifthe masses are taken to be the effective masses
of tile material being analyzed. The height of the confining potential is determined by
the band offsets, see Fig. 4.12.
Considersemiconductorquanturn dots fabricated from a two-dimensional heterostruc-
ture by using an etching process, or by colloidal synthesis, as illustrated in Figs. 7.12(b)
and 7.12(c). In these cases, depending on the heterostructure type. it is possible 10 realize
different types of carrier confinement. The classification of heterostructure types was
presented in Section 4.5. If the initial heterostructure is of type I, both electrons and
holes can be confined in the same quantum-dot stmcture.lfthe initial heterostructure is
of type II, only one type of carriers, either electrons or holes, can be confined.
Ifelectron motion is quantized in all three possible directions, we obtain a new physical
object, a macroatom. Questions concerning the usefulness ofsuch objects forapplications
naturally arise from the point of view of their electronic applications. A fundamental
question is the following: what is the current through a macroatom? A valid answer is
Ihal there exists the possibility of passing an electric current through an artificial atom
due to tU1lneling of electrons through quantum levels of the macroatom.
Importantly, that transport of charge occurs always in quanta of tile elementary charge
of the electron. This discreteness has 110 practical consequences for current flowing in
bulk materials, in two·dimcnsional structures, and even in quantum wires, because the
numbers of electrons transported are large. Charge transport in quantum-dot systems
is completely different. Indeed, it occurs as electron tunneling from a cathode through
the quantum dot to an anode. That is, the clearance space between the cathode and the
anode should be considered as a potential barrier in which a quantum dot is embedded.
In Section 3.3, while studying the tunneling effect, we found that the probability of
tunneling drastically depends on the thickness and height ofthe barner: for large enough
thickness and barrier height, tunneling is essentially suppressed. However, tunneling
through a discrete state localized inside the barrier has much higher probability. The
whole ttmneling process can be thought of as a sequential process: the capture of an
electron from the cathode to the localized state and then its emission to the anode. As a
result, the easiest way for the electrons to be transferred from the cathode to the anode
is for them to tunnel through the quantum dot. Then, an electron being captured to
the quantum dot blocks tunneling of other electrons. Electron transfer occurs in highly
correlated manner, one by one, at least at low temperatures. Examples of such "single-
electron" transport were discussed in Section 6.5. TIlUS, the electnc current through a
quantum dot occurs in the regime of single-electron transport. Devices that use such
transport will be discussed in the next chapter.
In fact, e1ectricalmcthods applied to quantum dots to realize useful devices are not
the only methods possible. The control of the electric current through the dots can also
be realized by means of light, sound waves, etc. Consider here optical control of the
dots and optoelectronic functions of zero-dimensional devices. The main peculiarities
of the optical properties of quantum dots arise due to electron and hole quantization.
In quantum dots fabricated using a type-I heterostructure, the carrier energies have the
form
£?D = E~D + En (" I. "2, 113). E~D = -E p (I1;. 11;, 11~).
Here, E~o is the fundamental bandgap of the material of the quantum dots. For a type-I
heterostructure, E~D is less than the bandgap E g of the surrounding material into which
the dots are embedded; en and £p depend on sets of three discrete quantum numbers,
{Ill. "1. II)} and {II;. ,,~.I1~}, for the electrons and holes, respectively. Specific depen-
dences of E?D and E~b on the quantum numbers are determined by the materials used
for quantum dots and their environment, the geometry of dots, etc. TIle model depen-
dences ofEqs. (3.51) and (3.53) can be used for estimation of the energy levels. Owing
to these discrete energy spectra, quantum dots interact primarily with photons of discrete
energies:
(7.10)
-~'\·:::::::::::::::::::::::·r",--E"
QD QW
, , , , ,
1.15 1.20 1.25 Energy (eV)
........... ..
.............. [
R x
• ••
GaAs
InGaAs
GnAs
lOOnm
Figure 7.13 A sketch of quantum dot confining potentials for electrons and holes. Peaks in optical
spectra correspond to phototransitions between quantized states of the electrons and holes in the
quantum dot. Re.printed with permission from 1. Tulkki and A. Heinamaki. "Confinement effect
in a quantum well dot induced by an InP stressor," Pllys. Rev. B. 52 (II), 8239 (1995). © 1995
by the American Physical Society.
where c is the velocity of light and A is the wavelength of the light. The different combi-
nations of quantum numbers tn I . n 2. 11 J J and {II; . ,,;. II;} give a series of optical spectral
lines, for which interaction between the dots and light is efficient. Importantly, the fact
that E2 D
< Eg implies that the light interacting with the dots is nol absorbed by the
surrounding material. These considerations are illustrated in Fig. 7.13. TIle potential
profiles of the conduction band, Ec• and the upper (heavy-hole) valence band, Ehh. are
depicted. The potential wells for electrons and holes represent the confinement poten-
tials of the quantum dol. Quantized levels for electrons and holes are shown; arrows
indicate the possible phototransitions between different quantized states. For each such
transition, a characteristic line of the optical spectrum arises. For the example of an
InGaAs quantum dot grown inside a GaAs quantum well. the quantum-dot spectmllines
arise in the region of photon energies from 1.1 to 1.3 eV, as presented in Fig. 7.13. The
width of observed spectral lines depends primarily on two factors. First, the spectral
lines arc broadened because of relaxation processes in the system. Second, it is typi-
cal that in experiments light interacts with numerous dots and there is some dispersion
of their sizes. This produces a dispersion in energy-level positions and additional line
broadening.
The optical control of the electric current flowing through a quantum dot can be
explained with the help of a device that can be called a single-quantum-dot photodiode.
A1a Sl.'1"= 927.35 nm T~ 4.2 K

25
<,20 927.62 11m
5
-"
Q)
15
927.1 11m 928.03 nm
....
~
10
"
u
-
0 928.45 nm
5
.A
0
..c:
0.. '-..I
0
0.4 0.6 0.8 1.0 1.2 1.4

Bias voltage (V)
Agure 7.15 PhOlocurrenl resonances for several different excitation wavelengths versus bias
vollagc. Reprinted with permission, from H. Kenner, S. Sufter et al., "Recent advances in
exciton-based quantum information processing in quantum dot nanostructurcs;' New J. Phys.• 7,
184 (2005). () lOP Publishing Limited.
electrons and holes inside the quantum well, which produces the measured pholocurrcnt.
As the applied bias increases, the energies arc shifted to smaller values and the resonance
wavelength increases. In Fig. 7. I5, spectra for excitation of the same ground state of the
dot for different biases are shown. The observed photocurrent spectra arc very narrow
because a single dot is involved. Spectral broadening becomes visible at high biases when
the electron and hole energy levels decay as a result of the increased rate of tunneling
from the dol.
A photocurrent excited resonantly with an electrically tunable optical resonance can
find numerous applications. Specifically, such a quanlum·dot photodiode facilitates opti.
cal manipulation of individual quantum states tunable with an electric bias. Such a
capability is necessary for devices required as components of systems used in quantum
information technology.
7.5 Closing remarks
In this chapter, we focussed on traditional quantum struct'ures that are already being
exploited in nanoelectronics and optoelectronics. These include quantum wells, wires,
and dots. These structures provide electron confinement in one, two, and three dimen·
sions, respectively. We found that confinement ofthe electrons in any direction forbidding
free propagation in this direction leads to quantization of the electron energy.
We showed that the confinement effecl can be very strong. In particular. for one orthe
most lIsed and practical cases - when electrons are confined at a heterointerface - we
found that the electrons are localized within a thin spatial layer of thickness ranging from
5 nm to 10 nm depending on the electron concentration. This confinement results in high

electron densities. For example, a surface concentration of confined two-dimensional
electrons of about 10 12 cm- 2 corresponds to a bulk concentration of from 10 18 cm- 3 to
2 x 10 18 cm- 3 , which arc typical values for heavily doped semiconductor bulk materials.
In quantum wells and wires, the carriers can be spatially separated from layers with
dopants, which generate free carriers. Such techniques lead to quantum structures exhibit-
ing large electron mobility relative to the case of bulk structures. Together with high and
controllable electron density, this enhanced mobility results in high electric current den-
sities and high speeds of operation.
Quantum dots possess totally quantized spectra for electrons and holes. Electron trans-
port via the dots can occur in the tunneling regime. Accordingly, the optical properties of
the dots differ considerably from those of bulk samples, as well as from those of quantum
wells and wires. Quantum-dot spectra consisl of a series of separate lines. The positions
of these spcctrallines can be controlled by dot dimensions and geometry. Quantum dots
find applications in optoelectronics.
For those who want to study traditional quantum structures in detail, we recommend
the following additional reading.
Results on electron quantization in Si021Si structures are presented in lhe review
T. Ando, A. B. Fowler, and F. Stern, "Electronic properties of two-dimensional sys-

tems," Rev. Mod. Phys., 54, 437 (1982).
Discussions of particular examples of the quantum wells and wires based on Ill-V
compounds can be found in the following references:
G. Bastard, Wave Mechanics Applied to Semicollductor Heterostructures (New York,

Halsted Press, 1988).
C. Weisbuch and B. Vinter, QU(ll/lIIm Semiconductor Structures (New York, Academic
Press, 1991).
V. V. Mitin, V. A. Kochelap, and M. A. Stroscio. QlI(llIlIIm Hetero.'l1ructures (New
Fabrication methods and properties of quantum dots are analyzed in the book
D. Bimberg, M. Gnmdmann, and N. N. Ledentsov, QlIal1lL1m DOl Helerostrllctllres

(Chichester, John Wiley & Sons. 1999).
7.6 Problems
1. Consider a quantum-well structure with the electron energy given by Eq. (6.27).
Assume that for the lowest two-dimensional subbands the intersubband distance Ll£21 =
£2 - £J is given. For T ~ 0, find the formula for the critical concentration lit of two-
dimensional electrons at which the second subband starts to be populated. Using el, e2,
and Llell for an infinitely deep quantum well. estimate the critical concentration, nt, for
a quantum well of width L = 10 mll.
7.6 Problems 239
2. To obtain characteristics of the electrons confined at an interface, one can use the
electron wavefunction of the lowest quantized state in the form of Eq. (7.9), which
depends solely on the parameter b. Apply the definition of the expectation value of a
physical quantity presented in Eq. (3.12) and find an expression for the average distance
ofthe electrons from the interface {z}. Calculate the mean square deviation ofthe location
ofc1cctrons from its average position; i.e., calculate {(z - (z})2).
3. Near the semiconductor hetcrojunction the electrons can be confined in the direction
normal 10 the heterojlmction. In Figs. 7.1 and 7.2, sketches of confining potentials are
presented. If the potential has high barriers, the lower energy levels may be studied with
some accuracy by applying the following "triangular" approximation:
V(x) = leFx. for x > O.

00. for x ~ O.
Here, x = 0 corresponds to the heterojunction position, and F can be interpreted as

the confining electrostatic field. For this triangular model of the quantum well, the
Schrodinger equation for the transverse component of the electronic wavefunction can
be solved exactly. The quantized electron energies are
where Pn are parameters defined by some algebraic equation. The lowest parameters are
known to be PI :=:::: 2.35 and P2 ~ 4.09. Using the effective massofGaAs, /II" = 0.067/110,
evaluate the positions of the lowest energy levels for the AIGaAs/GaAs heterojunction
as a function of the field F.
The electrostatic field F can be related easily to the electron concentration II g inside
the triangular well. Indeed, ionized (positive) impurities in the wide-bandgap part of the
hetcrojunction (x < 0) determine the confining field F. The number of these impurities
approximately equals the number of the electrons. Thus, in the region x > 0, Gauss'
law gives the field as F = 4Jrell g /E, where E is the dielectric constant. Assuming a
concentration of lis = 10 12 CI11- 2 and taking l = 13, estimate the field F and the energies
0$1 and 0$2. Compare the separation between the first two levels with the thermal energy
kB T at room temperature. Discuss the population of these levels by the electrons at T =

300 K.
4. In Section 4.4 on analyzing different types of crystals, we found that in silicon
there are six energy-equivalent minima (valleys). The electrons are mainly accumulated
near these minima. The minima are at the following crystalline directions: (100], (iOO].
[OIOJ. [oTO], [OOIJ, and [OOT] (see Table 4.4). Within these minima, the electron energy
spectrum is given by Eq. (4. I7) with effective mass tensors of the type of Eq. (4.19).
Let us consider a Si quantum-well layer oriented perpendicular to the (001] direction.
For free electrons collected in the pair of valleys along directions (001] and (00 i], the
electron Hamiltonian (kinetic energy) is
while for the other two pairs of valleys the Hamiltonians are
t,2 a2 tt 2 a2 t,2 a2
---- -------
21111 ax 2 2m[ ay 2 2m, az 2 '
and
t,2 a2 tt2 a2 t1 2 a2
-2m-ax-2 -21111
- flyl
- -2m,
- az
-l '
l
ll1esc Hamiltonians imply that the electron motion along the quantum-well layer and
that in the perpendicular direction arc completely independent. The wavefunctions and
quantized energies should have the form of Eq. (7.2).
Assume that the potential barriers of the quantum well are infinitely high. Employing
the results ofEqs. (3.27) and (3.28) for an infinitely deep potential, find energy subbands
forthe different electron valleys. Discuss the energy splitting ofthe valleys in the quantum
well. Using the data for 1111 and"" presented in Table 4.-.1, determine which pairofvalleys
will be of the lowest energy. In bulk Si crysl'als, lhe numbers of electrons in different
equivalent minima are equal. Can we expect an electron redistribution to be caused
by quantization in the well? Which valleys will be overpopulated and which will be
depleted?
5. A nanostructure consists oftwo "semi-infinite" hila-dimensional electron gases con-

nected through a quantum wire. The entire structure is fabricated from a single "infinite"
quantum-well layer. Does a potential barrier arise for electron motion from one IWo-
dimensional gas to the another? Explain why the barrier arises.
Using the adiabatic approach described by Eq. (6.124), discuss the barrier heights for
the electrons from different two-dimensional subbands.
Assume that the thickness of tile quantum-well layer fabricated from GaAs is 5 nm and
that the quantum-wire width is 5 nm. Estimate the height of this barrier for the electrons
populating the lowest hvo-dimensional subband.
6. In lll-V compound semiconductors the energies of heavy and light holes coincide
at zero waveveclor, as shown in Fig. 4.10. Energy levels exhibiting such a coincidence in
energy are known as degenerate energy levels. Quantization ofthe two types of holes lifts
this degeneracy. Using models wit h infinite-potential wa Jls for quantum wells, wires, and
dots, calculate the splitting between the lowest states of heavy and light holes in these
low-dimensional structures.
Obtain numerical estimates of this splitting for GaAs structures with the following
geometrical parameters: a quantum well of thickness 5 nm, a quantum wirc of cross-
section 5 11m x 5 nm, and a quantum dot of volume 5 run x 5111n x 5 nlll. Parameters of
heavy and light holes are given in Table 4.5.
7.6 Problems 241
7. It is known thai GaAs/AIGaAs helcrostructures are of type I. Thus, both electrons

and holes are confined in low-dimensional structures fabricated from these materials.
Consider GaAs quantum dots embedded in an AIGaAs matrix. Use the parameters of
electrons and of heavy and light holes presented in Table 4.5. Explain what kind afholes
forms the hole ground stale in a GaAs quantum dot.
To describe the quantum dots, apply the model of a potential box of Eq. (3.51). Using
Eq. (7.10), calculate the positions of spectral lines for quantum dols with the following
dimensions: (a) Lx = L y = L: = 5 11m and (b) L.t = 31l1n, L y = 5 nm, and L= = 7 nm.
The bandgap of GaAs is given in Table 4.5.
8 Nanostructure devices
8.1 Introduction
In previous chapters, fundamental physical processes on the nanoscale, analysis ofnano·

materials, and nanofabrication methods were all discussed extensively. The knowledge
gained in these previous discussions makes it possible to consider and analyze a vari-
ety of different nanostructurc devices. In this chapter, we consider electronic, optical,
and electromechanical devices. Some of these devices mimic well-known microelec-
tronic devices but with small dimensional scales. This approach facilitates applications
to devices with shorter response times and higher operational frequencies that operate
at lower working currents, dissipate less power. and exhibit other useful properties and
enhanced characteristics. Such examples include the field-effect transistors and bipolar
transistors considered in Sections 8.3 and 8.5.
On the other hand, new generations of the devices are based on new physical principles,
which can not be realized in microscale devices. Among these novel devices are the
resonant-tulUleling devices analyzed in Section 8.2, the hot-electron (ball istic) transistors
of Section 8.5, single-electron-transfer devices (Section 8.4), nanoelectromechanical
devices (Section 8.7), and quantum-dot cellular automata (Section 8.8).
As a whole, the ideas presented in this chapter provide an understanding of the future
development ofnanoelectronic and optoelectronic devices that may be realized through
the wide use of nanotechnology.
8.2 Resonant-tunneling diodes
Diodes or, in other words, two-terminal electrical devices, are the simplest active ele-
ments of electronic circuits. Some applications of diodes are based on their nonlinear
current-voltage characteristics. Another important capability required of diodes is their
operational speed. Such high-speed operation implies that the feature sizes of diodes
should be as small as possible.
In previous sections, we studied two types of short n+ -i- n+ diode, both with space-
charge·1 imited classical electron transport. These diodes used hOl11ostructures, i.e., it was
supposed that they are made frol11 the same material with nonuniform doping. Among
the cases studied, the highest speed is achieved for the ballistic diodes. According to
the classification given in Section 6.2 for the extreme scaling down of the diode size,
8.2 Resonant·tunneling diodes 243
(a) (b)
,, ,,
,, ,,
Vb , n+-GaAs Vb ,
,, ,,,
~s
,, · ,
· ·
L, AlxGul_.\·As AlsGaHAs . ·
Lb
·
I L, GaAs GaAs . ·I L
·
AlsGal_sAs
· ·
s ·
n+-GaAs
Figure 8.1 Layer designs for double-barrier resonant-tunneling structures. (a) Alternating layers:
n+ -dopcd GaAs (substratc and one of the contacts), undoped AIGaAs (barrier), undoped GaAs
(quantum well). undoped AlGaAs (barrier), n+ -doped GaAs (top contact). (b) The same as (a)
except for two additionalundopcd spacer layers of GaAs between the contacts and the barriers.
Vb is the barrier height, Lb the barrier thickness, and L .... the quantum well width.
quantwn transport must be used to describe the carriers. A very important example of
such nanoscale diodes is the so-called double-barrier resonant-llII1flelhrg diode. In this
section, we consider this type of nanoscale quantum diode.
The physics underlying the resonant-tunneling effect

We slart by considering a double-barrier heterostructure as an example of resonant-
tunneling diodes. Figure 8.1 depicts a sequence of layers in such a structure. The top
and bottom parts of the structure are doped regions, while the barriers and well layers
are undoped. Figure 8.1 (a) shows a specific structure in which a quantum-well layer of
GaAs is embedded between two AlxGal_xAs barrier layers. The top and bottom regions
ofdoped GaAs serve as contacts. A slightly different design is shown in Fig. 8.1 (b), where
two additional spacer layers separate the doped regions and the double-barrier part of
the structure. The purpose of these spacer layers is to prevent scattering of tunneling
electrons by impurities in the contact regions. The thicknesses of tile well, barriers, and
spacers may be varied substantially. Inside the quantum well. severn I quantized levels
can exist. In fact, these levels are quasi-bound states, because there is a small but finite
probability of the electron tunneling out of the well. Quannim-mechanical tunneling is
responsible for the finite lifetimes of the electrons in those levels and leads to some
broadening of the quantum-well states.
Thus, a resonant-tunneling diode can be thought of as a system with two contacts.
with three-dimensional electron states. and a quantum well, with a two-dimensional
electron system. These three subsystems are weakly coupled through tunneling. Energy-
band diagrams of the structure are presented in Fig. 8.2 for three different voltage biases.
Figure 8.2(a) corresponds to the equilibrium case, when no voltage is applied. In the well
under consideration, there exists at least one quasi-bound level; the caseofa single level of
energy, £1. is depicted in Fig. 8.2. Actually. £1 is the bottom ofthe lowest two-dimensional
(a) d
e
E.
E,
--•
(b)
E. e
- E.
E'; E""
,
e,
(c)
e
E.
,
£~
Agure 8.2 Energy band diagrams of double-barrier resonanHunncling structures:

(a) equilibrium conditions: (b) in-resonance: and (c) ofT-resonance. The electron potential
energy and populations orthe energy band in contact n+ -regions are sketched in the left and
right pans of the figure. E:'" and E~ denote the positions of the bonoms oflhe conduction
bands under electric bias.
sllbband because there exists in-plane free electron motion. The parameters of the diode
are chosen 10 be such thai, in the non-biased state of Ihe diode, Ihe quasi-bound level,
E1, lies above Ihe fermi energy, EF, in the contacts, as in Fig. 8,2(3). By applying a
voltage bias to the contacts, one can produce a downward shift of the level in the well.
For electrons with arbitrary energies. the probability of tunneling Ihrough the double-
barrier structure is very small. The structure is designed to prevent the thermal transfer
of electrons over the barriers. Therefore, the only situation favorable for transmission
of electrons through the structure is when the quasi-bound level lies below the Fermi
energy, EF, but above the conduction-band bottoms of the contacts. In this case, those
electrons from the emitter (the contact on the letl) whose kinetic energy of the in-plane
(perpendicular) motion, E1. = ftl~/(2m"), coincides with EI arc transmitted through
the structure with finite probability. This is the so-called resonant-tunneling process,
which has the important attribute of exhibiting negative differential resistance.
Before analyzing this effect, we recall that the electrical properties of a simple con-
ductor with a linear current-voltage characteristic are characterized by a resistance, R,
according to the formula
J=¢>./R.
where 4>0 is lhe appl ied voltage and I is the current. I r a conductor has a nonlinear
current-voltage characteristic. one can introduce the so-called differential resistance.
Rd :
Rd= ( -dJ )-'

d¢>
8.2 Resonant~tunneling diodes 245
(d)
,
(a)
J 4t(C)
o
Figure 8.3 Resonant-tunneling diode current-voltage characteristic: the portions labeled by
(a), (b). and (c) correspond to the physical situations illustrated in the previous figure. Portion
(d) shows what happens with the current when the second quasi-bound state enters into
resonance with the electrons from the emiuer.
The term negative differential resistance is used to denote that Rd < 0; this corre-
sponds to the unusual effect of a decrease of the current when the applied voltage
Increases.
The following qualitative model ofresonallt tunneling explains the appearance of neg-
ative differential resistance. Before the level of the well reaches the resonance position,
the current through the diode is very small because it is controlled by non-resonant tWI-
neling and by transport over barriers, both of which have low probabilities of occurring.
When the bias corresponds to the case ofresollance energies, as depicted in Fig. !t2(b),
the transmission coefficient and the electric current througb tbe diode increase sharply.
Further increase in the current with increasing voltage bias continues until the resonance
level passes the bottom of the emitter's conduction band. There are no electrons (tbe
conduction band is above the Fermi level, EF). as shown in Fig. 8.2(c), to tunnel reso-
nantly and the current decreases in spite orthe increase in voltage bias. Consequently,
the current-voltage characteristic of the struchlre contains a segment that exhibits neg-
ative djfferential resistance. The general form of the current-voltage characteristic is
presented in Fig. 8.3. At large bias, further increase of the current can be realized either
by shifting other quasi-bolUld states so that they are resonant with electron energies in
the eminer, or by substantial evolution of the potential profile and non-resonant transport
through and over the barrier.
Quantitative characteristics of the resonant-tunneling effect

The resonant transmission of the electrons through the double-barrier diode can be pro-
vided by physically different processes. Conceptually, the simplest way is direct quantum-
mechanical tunneling, which corresponds to the coherent tllflneling process.
Coherent tunneling
In this case, the electron is characterized by a single wavefunction throughout the whole
quantum structure. Traditionally, the cathode and the anode of the diode arc referred to
as the emiuer and collector, respectively. An electron entering the structure from the
1.0 .....~......,..~~.,....,.,~
0.9 T (EU ~ _-,.,,-1__._,....
1+ (EtE
0.8
0.7
£a
1
26£= 80 meV
~ 0.6
~ 0.5 -----------------I-H-l'H lOy, = 0.1 eV
e-.; 0.4
0.3
0.2
0.0 ~ ==~~~~~=:J
0.1 to:::; ~
-0.2 -0.1 0.0 0.1 0.2 0.3 0.4
£1- (eV)
Agure 8.4 The full width, 2 Ii. E. of the half-maximum of the transmission peak.
emitter is described by the wavefunction 1Jtcrt1(r. =), which has the foml of an incident
plane wave,
and an outgoing wave that has passed through the structure,

1Jtcrt1(r. z) = BCmei(krHi':);
here, k and r are two-dimensional in-plane vectors, and = is directed perpendicular

to the layers. The quantities k~m and k~ol are the :-components of the wavevector in the
emitter and collector, respectively. In an ideal layered structure, the vector k is conserved
because there are no forces acting in plane directions; thus, k is the same for incident
and outgoing waves. The relation of these wavcvcctors, the corresponding energies, E1:
and E1ol • and the applied voltage. 4>0, can be obtained from the energy conservation
law:
h'(k~m)' h'(k'·')'
E.1~ - ECol
.1 =
- - - - = e <l>o·
2m- 2m-
The coefficients Ami and Bem can be found by solving the Schrodinger equation with
the potential corresponding to the double-barrier structure under voltage bias. The ratio
of the coefficients BmI and Ami defines the transmission coefficient:
IB~(E.LlI'
T(E ) - '-=~",
~ - iA,m(E~)I'
Inst'ead ofsolving the SchrOdingerequation. here we give a simplified formula for T(E.l).
which allows us to introduce the basic parameters of the resonant-tunneling effect.
8.2 Resonant-tunneling diodes 247
As follows from the qualitative analysis presented in the previous subsection, T(El.)
has 10 have sharp peaks in the vicinity of resonance energies, En. For a structure with
synunetric barriers, we can approximate these peaks with
(8.1 )
where 2 t!J.E is the full width of the half maximum of the transmission peak. Equa-
lion (8.1) assumes that the probability of tunneling through the structure is unity when
the electron energy, £.1, exactly coincides wilh a quasi-bound state, En. In Fig. 8.4, T(El.)
is plotted for several magnitudes of 6£. One can sec that the tUlmcling transmission
coefficient can be a very sharp flmctioll of the electron energy.
Besides the transmission coefficient, it is convenient to introduce the probability of
tunneling of the electrons out of the well per second, r; it is determined by 6.£:
r = "£1".
Accordjng to the uncertainty relation we can estimate the lifetimc, T, of thc clectrons on
the quasi-bound level between the barriers as
r = 1/(2r).
The tunneling probability per unit time can be expressed as the product of the attempt
rate, 11;/(2L w ), and the probability of tunneling through a single barrier with one
attempt, T:
where L w is the thickness of the well and 11; is the transverse velocity of the electron in
the well, which can be estimated from the relationship
For the purpose of making illustrative numerical estimates, we choose the following sci
of parameters: the energy of the first level, E:I, is 50 meV, the height and thickness of
• •
the barriers are Vb = 300 meV and Lb = 40 A, and the well width is L w = 100 A. For
this structure, we obtailll1; = 5 x 10 7 cms- I and r = 1011 S-I. Hence, the lifetime of
an electron in the well equals t = 5 X 10- 12 s. Both T and r depend critically on the
height and thickness of the harriers.
The lifetime of the quasi-bound level depends on heterostructure parameters and
material combinations. In Fig. 8.5, the lifetimes of the quasi.bound states are depicted
for five particular double-barrier structures. The structures are assumed 10 be made of
GaAslAIGaAs,lnGaA AlAs, InGaAslGaAs, and InAs/AISb. The widthsofthe quantum
• • •
wells are equal to 46 A, while the barriers have thicknesses ranging from lOA to 60 A.
The heights of the barriers vary from 0.3 cV to 1.2 eV and are marked on the lines.
Figure 8.5 shows that the lifetime can vary over a very wide range.
/
10 / /
/ /
- /"'- 1.2''Y
, /' /
/ /7
• V.. -O.3cV
~
/'
J'
/ '-
,
.....••-
a b /0 /0 d ....
"b""lcV./ / /' ••••••••
" . '/ ""'..,.-!,~"~l~l_-r_b~,~rr¥;~,,~,_
"'" I . .... -a GoAs AlAs
I / ..... -i'+'--::;;;:-+--""7"--~
/ • • b hI.gJJG~.r\s AlAs
I
'/ /
....
••••••• _'7t"..;G~"';;::'-.-:fA~[·~·f,G~"""!~As,,,
I G
.' d n"!l a....rI\S GaAs
c InAs AISb
0.01 0~~~1~--=-2~~3~........L4::::::~5::::::~6===:J7
Barricr thickness (11m)
Figure 8.5 The lifetime of the first quasi-bound state in n-type double-barrier structures made
from the five material systems presented in the inset. In each structure. the quantum well width
is fixed at 46 . The energy heights of the barriers are indicated next 10 the curves. Afl'er
T. C. L. G. Sollner et 01., "High· frequency applications of resonant·tunneling devices:' in
Electrollic PlVfJerties ofM/lltilayel:v {lilt! LOIr-Dimellsional Structllre.v. J. M. Chamberlain et al.
(Eds). (New York. Plenum, 1990), pp. 283-296.
For an asymmetric double-barrier structure, we define two different transmission coef-

ficients for the left-hand, Ii, and right-hand Tn barriers, respectively, so that the total
transmission coefficient, T, can be approximated by
I
(8.2)
where 6E = hr and r :::; (7i + Tr)v:/(2L w ). For asymmctric barriers, the maximum
transmission at El. = E" is less than unity. It is important to recognize that these results
arc obtained under the assumption thai the electrons transit through the system without
phase-changing scattering, so that El. remains constant.
Using the parameters introduced for the double-barrier resonant-tunneling structures
and assuming low temperatures, it is possible to obtain the estimates for lhe current-
voltage characteristics sketched in Fig. 8.3.
The formulas (8.1) and (8.2) describe the transmission ofan electron with fixed energy
El.' In order to calculate the electric current, we should take into account all of the
electrons which tunnel from lhe emitter and collector. We suppose that in the contact
regions - heavily doped regions - the thermal equilibrium for electrons is established in
a very short time. Therefore, to a reasonable degree of approximation we can aSSlUllC
that the Fermi distribution gives the distribution of electrons entering the double-barrier
part of the structure. If a voltage bias. <1>0, is applied, the difference between the Fermi
energies of tile emitter, E~m. and collector, E~()l, is

001
F - EF
E'" -- e<l> o·
In deriving Ihe Landauer formula in Section 6.5, we obtained the equation for the net
current in a quantum device:
l=rm_r:01 •
where, in terms of the present analysis, we introduced the current from the emitter to the
collector, fern, and the current from the collector to the emitter, I tol , These currents may
be expressed in terms of !.he electron concentrations in the two electrodes, ncm (E..1) and
"COI(£.d:
lelMOI = 2;1i JdE.l T(E.dncrn.eol(£J.). (8.3)
Here, the integration runs over the energies £1. above the bottom of the conduction
band in the emitter and collector, respectively. The concentrations, "em and "cOl, are
given by Eq. (7.6). Let us consider low temperatures, at which the expressions for the
concentrations arc simpler. Then, instead ofEq. (7.6) we obtain, for example,
em •
n (£1.) = !!!...,(£F - £1.). (8.4)
rrh-
The electron transverse energy is limited by £gm < E1. < £~.n\ otherwise the sheet
density of the tunneling electrons would be equal to zero. We get the following expression
for the current:
e 0 < E '"'

F' (8.5)
This formula takes into account the fact that, because all states in the collector with
energy £1. < E m - e<t>o are occupied, tunneling is possible only for those electrons with
r
energies satisfying the following conditions: £Frn > £1. > EFrn - e<1>o. For e<t>o > E~.rn
all emitter electrons can tlmnelthrough the barriers and we have
1=
em·
"
21T 11
lE-
E
?
dE~(E~m-E~)T(E~). e<l>o> qm (8.6)
•
We suppose that the Lorentzians in Eqs. (8.1) and (8.2) are narrow relative to £F nl
. 111en,
the integration over £1. in Eqs. (8.5) and (8.6) is straightforward and yields
em·lJ= ern 1j7;.
I = 2rrf,'L
w
(E, - E,,(o)) 7j + T,' (8.7)
where we have used Eq. (8.2) for asymmetric barriers. This equation represents the
tunneling current through the bound state with energy 6//(<1». which depends on the
voltage bias as ilJustrated previously by the example of Fig. 8.3(b). A peak value of
the currcllI is reached when the bound-state energy is resonant with the bottom of the
conduction band in the emitter £~m:
(8.8)
Table 8.1 Resonance width and collision broadening of levelE! of A1 0.3Gao. lAs1GaAs
double·barrier structures
r,lf sc
•
L•. (A) L, (Al r, (meV) atT=300K atT=200K al T=70K
50 70 1.3 X 10- 2 6 X 10- 3 1.9 X 10-2 2.6 X 10- 1

50 50 1.5 X 10- 1 7.5 X 10-2 8.3 X 10- 1 3.1
50 30 1.76 0.88 1.3 3.6
20 50 6.03 3.02 4.56 124
When the bound-state energy, En. falls below the bottom of the conduction band in the
emjttcr, the current drops rapidly to the value determined by off-resonance tunneling
processes. This vallie can be estimated, if we assume that off·resonancc transmission
coefficients are constant:
(8.9)
Since for the transmission coefficients the inequality 7i. Tr « I is valid, the resonant
current ofEq. (ItS) is much greater than the off-resonance current ofEq. (8.9). The results
obtained for the coherent mechanism of tunneling through a double-barrier structure
support the qualitative discussion given at the very beginning of this section.
Sequential tunneling
Another process responsible for the resonant-tunneling effect is the so-called sequential
tunneling process. In the sequential tunneling scheme. electron transmission through the
structure is regarded - somewhat artificially - as two successive transitions: first from
the emitter to the quantum well and then from the well to the collector. 11 is important
to highlight the main difference between the previously studied coherent mechanism
and the sequential mechanism of resonant tunneling. The first mechanism excludes any
electron collisions during Ihe transition from the eminer to the collector. The second sce-
nario applies even when there is electron scattering inside the quantum well. Although
the coherent and sequential processes result in the same behavior of the double-barrier
resonant struchlrcs, it is possible and instructive to separate and compare these pro-
cesses. In the previous discussion, we defined a broadening, r. for the quasi-bound
state due to the tunneling process. If both the processes of tlmneling and scattering
inside the well occur, the width of the quasi-bound state increases. Let us introduce
the width of the quasi-bound state, r r, as the full width of the half maximum of the
transmission peak. Lei the collision broadening of this state be r lie' From the previ-
ous discussion, we conclude that coherent tunneling dominates if rr > r $C, whereas
the sequcntial processes dominate if rr < r sc. Table 8.1 illustrates different regimes
of tunneling for several specific Alo.3Gao.7As/GaAs double-barrier structures at zcro
bias. In Ihis table, L . . . and L b are the thickncsses of the wclls and barriers, respectively.
TIle ratio rrl r sc is caJculatcd by using the estimate r sc = tIlT:, where T: is the scatter-
ing time. From the results of Chapter 6, this time may be either calculated from first
8.2 Resonant-tunneling diodes 251
15
100
300K ,
50
o (a)
-50
-100
_150'1-----------
150 80 K
100
50
o (b)
-50
-100
_ 150 1------"-----------.J
-1.0 () 1.0
Bias voltage (V)
Agure 8.6 Measured current-voltage characteristics of a symmetric Alo4slno.52As/Gao..ulllo.BAs
double-barrier resonant-tunneling structure al 300 K (n) and 80 K (b). Aftcr Fig. I, F. Capasso,
S. Sen et af.. "Quantum-effect devices," in S. M. Szc (Ed.), High·Speed Semicomillctor Del'ices
(New York. Wiley. 1990). Reprinted with permission of John Wiley & Sons. Inc.
principles or deduced from experimental mobility measurements. For Table 8.1, the fol-
lowing values arc assumed: when T = 300 K, J.L = 7000 cm 2 V-I 5- 1, and r ::::: 3 x 10- 13
s; when T = 200 K, /.L = 2 X 104 cm I V-I 5- 1 • and r = 10- 12 s; and when T = 70 K,
J1, ~ lOS cm 2 V- 1 5- 1, and r::: 5 x 10- 12 s. If rr/r iC > I, clastic collisions can be
neglected and resonant tunneling is highly coherent. From Table 8.1 one can see that
this type of tunneling is typical for the case of low temperatures and thin barriers. In
the opposite limit, rTf r sc: « I, the tunneling processes arc more likely to be sequen-
tial. II is important to emphasize that the scattering affects also the magnitude of the
transmission coefficient. The maximum of the transmission probability decreases by the
factor rr/Crr + r sc:) when scattering takes place. This explains why resonant tunneling
is washed out at high temperatures and in structures with defects and impurities.
Negative differential resistance under resonant tunneling

As we have already seen from qualitative discussions, resonant-tunneling structures man-
ifest strongly nonlinear current-voltage characteristics. In particular, it is very impor-
tant that there is a portion of these characteristics for which the current decreases
while the voltage increases. Typical current-voltage characteristics for a double-barrier
structure of AllnAslGalnAs are presented in Fig. 8.6. The results are shown for two
4.0
3.5
3.0
T~ 300 K
InAslA1Sb
\
,
~
N I
I J,IJ" = 3.3
~
~
~
"u
2.5
2.0
\ I
,
0 GaAsiAIAs \
~ 1.5
\..
-,
1.0
J,IJ, \1.4
.l,
I
- ,
InGaAs/AIAs
0.5 \
0.0
0.0 0.2 0.4 0.6 0.8
....
1.0
-JplJ" \ 12
1.2 1.4
~
1.6 1.8 2.0

Voltage (V)
FIgure 8.7 The room·temperature current density versus voltage characteristics and peak.!(}-
valley mtios, Jpl lv, for high-speed double-barrier RTDs made from three different nUllerial
systems: GaAslAIAs. InAslA1Sb, and InGaAslA1As. After E. R. Brown, "Resonant tunneling in
high-speed double barrier diodes:' in Ho' Carrier,f ill Semiconductor NlII/Oslruclw'P,f: Physics
and Applications, 1. Shah (Ed.) (Boston, MA, AT&T and Academic, 1992), pp. 469-498.
temperatures. T = 80 K and T = 300 K. The structure is symmetric (similar barriers,

equal thicknesses of spacers, and doping of contacts); thus, there is au almost antisym-
metric current-voltage characteristic. For the lower temperature, there is clearly seen a
portion with almost zero current at finite voltage biases. which corresponds to the posi-
tion ofthe resonant quasi-bound level above the Fermi level of the emitter and very small
non-resonance and overbarrier currents. When the level is shifted below the bottom of
the emitter's conduction band (<1>0 ~ ±0.75 V), the current drops to almost zero values.
At room temperature, small ovcrbarrier currents exist at any finite voltage bias and the
current drop is also considerably less than in the former case. An important parameter
for characterizing N-type negative differential resistance is the ratio of the maximum
and minimum values of the current - the so-called peak-Io-valley ralio. Achieving large
peak-to-valley ratios greatly enhances the possibility of applications utilizing negative
differential resistance. For the case presented in Fig. 8.6, the peak-to-valley ratio is 15
at 80 K. At room temperature, the ratio decreases to about 4, but still remains sufficient
for applications. The peak-to-valley mtio depends not only on the physical nature of
the negative differential resistance, but also on many technological and design factors.
Although the development of the double-barrier system is still in progress, for the most
optimized and perfect structures, a peak-to-valley ratio of about 20 or even higher may
be achieved at room temperature.
Comparisons of the current-voltage characteristics for various double-barrier struc-
tures and the peak-to-valley ratios obtained for T = 300 K arc presented in Fig. 8.7. Three
types of heterostructures currently used for resonant-tunneling diodes - GaAs/AIAs.
InGaAslAIAs, and InAs/AISb - are characterized by current densities exceeding

105 A cm- 2 and exhibit negative conductance al room temperature. The largest peak-to-
valley ratio is rcached for InGaAs/AIAs structures (above 10 at room temperature), but
the highest current densities are, typically, for InAs/A1Sb, Jm :::::: 4 x lOs Acm- 2. For
GaAs/AIAs structures, these parameters arc relatively modest: the peak-ta-valley ratio
is about 1.4 and J m :::::: 1.5 x lOs A cm- 2 .
Another important parameter of any system exhibiting negative differential resis-
tance is the characteristic time of the processes responsible for the negative differential
resistance. TIlis time determines the physical upper frequency limit at which the nega-
tive differential resistance disappears. For the previously considered resonant-tunneling
device, the frequency limit is not easy to estimate because this type of electron transport
has no classical analoguc. Careful analysis of the frequcncy properties of tile tunneling
can be done only by numerical self-consistent calculations involving the time-dependent
Schrodingcr equation, the kinctic equations describing processes in the contacts, and
Poisson s equation. This complex problem has not yet been solvcd. However, by modu-
lating the barricrs by a small time-dependent voltage and examining the time-dependent
response of the system, it has been shown that the characteristic time of the tunneling
processes can be estimated by the following formula:
,' dz 2h d 2h
~--"'·o
1 j2m[V(z) - E~l
+-=-+-.
r ", r
(8.) 0)
where d is the total thickness of the two barriers and the well, and tJ g can be interpreted
as the electron group velocity. If the perpendicular energy of the incident electron, E.1,
is much greater than the resonance width, r, it can be shown that the total transit time
through the structure is given approximately by Tlr . The first term represents the semi-
classical transit time across the structure and the second term is the so-called pIlOse time.
•
For a typical example of a symmetric resonant-tulmcling structurc with 17-A-thick
•
AlAs barriers and a 4S-A-thick well, the quasi-bound level has an energy el ::::: 0.13 eV;
hence, 21iJ r = 0.45 x 10- 12 s. For a drift velocity tJd :::. 10 7 em s-I, the first term gives
only 0.8 x 10- 13 s. Thus, Tlr::::: 0.5 x 10- 12 s. Hence these quantum devices are ultra-
fast, with response times in the subpicosecond range.
A resonant-tunneling diode as a microwave oscillator

The application of tile resonant-tunneling effect in high-frequency oscillators is based on
the existence of the negative differential resistance. In order to review the principles of
using negative differential resistance for obtaining electrical oscillations, we consider the
simplest electric circuit containing a resistance. Rd , a capacitance, C, and an inductance,
L, as represented by Fig. 8.8(a). Let us introduce the resistance Rd as the ratio
where 6./ is the change in the current through the resistor when the voltage drop is
changed by 6.<J>. Thus, in fact, Rd is the dijJerelllia/ resistal1ce that can have both positive
(b)
(a)
,
~
N
Rd E
u 10'
~
'--
;>,
C L. "
u
~
10'
GaAslAIAs
Cl
~
u
3
0 10'
0-
-+
100 L-~~~.....,._~~...J
10' 1~ 1~
Frequency (GHz)
Figure 8.8 (a) The simplest electric circuit. (b) The generated microwave power per unit
cross-section of double-barrier resonant-lUnneling diodes as a function of the frequency for the
same thn.:e devices as in Fig. 8.7. (b) After E. R. Brown, "Resonant tunneling in high-speed
double barrier diodes," in Hot Carriers ;n Semiconductor NmlO.~t"lIctures: P"y.~ics and
Applications. 1. Shah (Ed.), (Boston, MA. AT&T and Academic, 1992). pp. 469-498.
and negative sign. If the alternating current, T, and voltage,~, applied to the circuit arc
and
one has the following relation between the magnitudes 10 and <1>0:
10 = z:w) <1>0. Z(w) = Ro - ;(wL - wlc). (8.11)
where Z(w) is the impedance of the circuit. The impedance - complex resistance -
includes the active resistance, Ril.
Even if<J>o = 0, oscillations can exist in the circuit at frequencies for which Z(w c ) = 0:
LC
I (R2LO
)' (8.12)
For LIC > (R d /2l and for a positive resistance Rd , there arc oscillations with the
frequency j I /(LC) - [R o/(2L )]' and damping characterized by
y = -Im{w,} = Ro/(2L).
Here, y is positive and Im{w} denotes the imaginary part of We'

Ifthe real part of the impedance is negative,
Re{Z(w)} = Ro < O. (8.13)

8.3 Field-effect transistors 255
it follows that y is negative and the steady state of the circuit is unstable with respect to
the generation of voltage oscillations. Thus, if we want to realize voltage oscillations at
a desired frequency, w, the differential resistance at this frequency should be negative.
The fact that double-barrier resonant-tunneling structures have extremely small electron
transit times means that these devices exhibit negative differential resistance up to ultra-
high frequencies. In Fig. 8.8(b), the generated microwave power per unit cross-section
of double-barrier resonant-tunneling djodes is presented as a function of the frequency
for the same three devices as were considered in Fig. 8.7. The generated power decreases
with frequency as a result of the previously discussed finite transit time of tunneling
electrons. Absolute values of the power correlate with the current densities achieved for
these devices. It is seen that oscillations up to I THz = 1000 GHz(i.e., 10 12 oscillations
per second) have been reached for nanoscale quantum devices of this type.
Finally, the resonant-tunneling structures are the simplest quantum devices. They
exhibit strongly nonlinear current-voltage characteristics with negative differential resis-
tance. Since these devices have nanoscale sizes, they exhibit extremely short transit
times for carrier transport through the structures. These properties allow one to exploit
resonant-tunneling structures for the generation of ultra-high- frequency electromagnetic
oscillations. Indeed, an oscillation frequency of about I THz has already been reached
in resonant-tunneling diodes.
8.3 Field-effect transistors
The previously analyzed two-terminal devices - diodes - are the simplest electronic
devices, for which the current is controlled by the diode bias and vice versa. A use-
ful function can be performed mainly due to nonlinearity of current-voltage depen-
dences. In contrast, in three-terminal devices known as transistors there exists the pos-
sibility of controlling the current through two electrodes by varying the voltage or the
current through a third electrode. Depending on the principle of operation, transistors
can be associated with one or other of two large classes: (i)jield-effecttransislOrs and
(ii) potential-effect transistors.
Devices controlled by the field effect

Devices ofthe first group arejield-effect or voltage-comrolled devices. A common feature
of these devices is that a voltage is applied to a controlling electrode - a gate - which is
capacitively coupled to the active region of the device. A capacitive coupling means that
by applying a voltage to the gate one creates a transverse electric field in the conducting
channel, but no useful current flows through the gate. The gate electrode is spatially and
electrically separated from carriers in the active region by an insulator or a depletion
region (a region where electrons are absent). The gate electrode controls the resistance
of the active region and, consequently, the current between two other terminals, which
are known as the source and the drain.
(0) (b)
I
-- n --- n
Semi-insulating Semi-insulating
substrate substmte
Figure 8.9 The MESFET model. The conducting channel of width / is supposed to be grown on a
semi-insulating substrate. The length of the channel is L.~. The depleted region. Id(X), under the
gate, G. is marked by a shadow region. In case (a) there is no voltage bias applied between the
source, S. and the drain. 0; in case (b) a positive vohagc is applied to the drain. There arc no
electrons in the depiction region.
There exist severnl different types offield-effect transistor (FET). Before discussing
the most important typesofFET, including those based on nanostructures, we will explain
the operation principles by using the simplest FET, the so-called metal-semiconductor
FET(MESFET).
Figure 8.9 presents a simple model for a field-effect transistor. It is assumed that the
active region of the device is made of an n channel, which can be fabricated, for example,
by homogeneous doping. The source and drnin are heavily doped n+ regions, which are
assumed to serve as Ohmic contacts. The gate upon the active layer forms a so-called
Schotl/..y barrier, which produces a depletion region, as discussed in Section 7.2 (see
the subsection on the control of charge trnnsfer). The formation of the depletion region
under the gate is illustrated by Fig. 8.9. Such a design is typical for commonly used
semiconducting materials such as GaAs and InP. From the bottom, the active region is
restricted by a non-conducting (semi-insulating) substrate.
If no voltage is applied to the contacts, the depletion region is almost uniform along
the active region, as shown in Fig. 8.9(a). The characteristics of the depletion region are
determined by the built-in (Schonky) voltage. If a negative voltage is applied to the gate
and there is still no voltage between the source and drain, the depletion region extends
farther into the active region and decreases the width of the channel. At some voltage
the channel is complele~y pinched off. Let the gate voltage be fixed and a small positive
voltage be applied to the drain. A current will flow through the channel in the Ohmic
linear regime. If the drain voltage is increased further, it will affect the distribution of
the potential in the device: the width of the depletion region increases near the drain end
of the channel, as shown in Fig. 8.9(b), as a result of the increasing potential difference
between the gate and the drain. At a certain drain voltage, the channel begins to pinch off
at the drain end. At this voltage the current saturates and remains nearly constant with
further increase in the drain voltage. This behavior is illustrated in Fig. 8.10, where typical
current-voltage characteristics ofa MESFET arc presented for various gate voltages. It
is clear that. if the gate voltage is negative, the channel becomes narrow and the current
is suppressed, whereas, for positive gate voltage, the depletion region under the gate
locus for satumtion

drain \ooltage <1>0...
I B~akdown
Salumtion ll:morl
Linear
regIOn
,
....._;;;;;;::.:;:;;;;.~~:;;;;...1 regloo
.-E=
,, >0
G - 0
Cl
<0
o "',
Drain volt'age o
Agure 8.10 The drain-current versus drain-voltage characteristics for the MESFET 011 different
gale voltages. The linear and saturation regions are indicated. The breakdown effect occurs at
high voltages 4>9 and is presented schematically.
Field-effect transistors (FETs)
Homogenous FETs MOSFETs Heterostructure FETs

Novel Devices
Ir--~"'-I
VMT QUIT
Figure 8.11 The family tree of FETs consists of three: groups of conventional fETs thaI have
much in common in their technologies, properties. and applications and new groups of novel
devices. The abbreviations used in this chari are defined as follows: MOSFET mctal-oxide-
semiconductor FET: VMT. velocity-modulation transistor; and QUIT. quantum-intcrferencc
transistor.
becomes narrower, the channel opens wider, and the current has to be larger. Thus, in
comparison with a diode, the more complex configuration of the electrostatic potential
in a FET gives rise to strongly nonlinear source-d.rain current-vohage characteristics.
Qualitatively. two operational regimes are possible, the linear region and the saturation
region, as indicated in Fig. 8.10. For both regimes, there is effective current control by
the gate voltage. At very large source-drain voltages, electric breakdown can occur.
In summary, the phenomenon of controlling the resistance of the conduction channel
by an external voltage (by a field) is the basic principle of any FET.
The FET-family devices

The FET family can be classified as shown in Fig. 8.11. where we give also short expla-
nations for the abbreviations used to denote these FETs. The FET family of transistors
SiD,
Source Gate Drain
n+
../ -----1
n-channel
p-type Si substrate
--
FIgure 8.12 A sketch of an Si-MOSFET.
is divided into three large groups of devices used in practice, as well as a group of
potentially new devices based on recent developments. The first group includes the vari-
ations of metal-semiconductor FETs discussed previously. The second group consists
of metal-oxide-semiconductor FETs (MOSFETs). In contrast to the previously con-
sidered MESFETs, in MOSFETs, the gate and conducting channel are separated by a
dielectric (an oxide) instead ora Schottky barrier. The MOSFETs are typically fabricated
by using a mctal-oxide-silicon structure, with the oxide layer being silicon dioxide. In
Fig. 8.12 we present a sketch of a Si MOSFET. It should be said that Si MOSFETs are
microelectronic devices of the utmost importance. A third group of FETs in Fig. 8.11
incorporates heterostructures and differs in the design of the doping and conducting
channel. For example, the donor-layer heterostructure FET (I-IFET) subgroup has one or
several doped layers, which supply electrons to the conducting channel. If the channel
is formed on a single heterointerface (see Section 7.2) using a single donor layer the
device is called a modulation-doped FET (MODFET). Such MODFETs have the largest
electron and hole mobilities. That is why they arc also called high-electron-mobility
transistors (HEMTs). Examples of material systems used for I-IFETs are as follows:
n+ -AIGaAs/GaAs (unstnlined) heterostructures, strained layer n+-lnGaAs channels on
GaAs, and n+ -lnAIAsflnGaAs heterojunctions grown on InP. Although the latter two
structures arc more complex, they offer better transport parameters in the conducting
channel.
Figure 8.13(a) shows schematically one of the variants of a HFET - a recessed-gate
n+ -AIGaAs/GaAs I-IFET. The device is fabricated on a semi-insulating GaAs substrate
upon which an undoped buffer layer of GaAs is grown. A thin undoped AIGaAs spacer
layer and an n-dopcd AIGaAs layer are grown on the buffer layer. Source, drain, and gate
electrodes are fabricated on the top of the structure. Under the electrodes, two heavily
n+ -doped regions serve as the contacts to the two-dimensional electron gas formed under
the heterojunction. Such a two-dimensional electron-gas channel was studied in Section
7.2 in detail. The gate length varies from 1.0 IJm to 0.1 1J1n, or less, depending on the
speed needed for applications.
Source Drain
A;
,
I~~
I I
G"",
I I
AIGaAsn+
(a) ., AIGaAs undopeJ
0."" "'...,.. I
-WEG
z
o.Aa w4w :
I
n,
V(z)
(b)
EF
'-' z
:soA
,
,,
(c)
r-iI
,, ,
~•
Figure 8.13 A recessed-gate n+-AIGaAs FET. (a) The cross-seclion of the HFET. (b) The energy
band diagnun along the line A-B; the Fermi level, E F , and electron confined states arc shown.
In the donor layer, there are two depletion regions for a thick barrier: one is due fO the Schottky
built-in voltage; the second is due to the conduction-band offset at the junction. (c) The
sequence oflayers along the line A-B.
The dimensions in the vertical directions can be different. This facilitates the fab-
rication of two types of HFET: normally turned-on and normally turned-off devices.
(Both types of heterostructures were studied in Section 7.2.) Figure 8.13(b) illustrates
the energy-band coordinate dependence for the cross-section of the structure under the
gate. The major element affecting the physical picture is the Schottl}' depletion layer
under the gate. The depletion layer is different for different thicknesses of the barrier
layer. For a barrier layer of small thickness, the Schottky barrier depletes even the quan-
tum well. leading to the normally turned-off state in the absence of the gate voltage.
For larger barrier thicknesses, the Schottky barrier does not prevent the formation of
the conducting channel, and we obtain the normally turned-on stale of the HFET as
illustrated in Figure 8. I3(b).
Let us briefly review the heterostructures employed for FETs based on Si technology.
Although many combinations of materials are being investigated with the goal of fab-
ricating high-quality Si-based heterostructures, the Si/SiGe system is the most studjcd
and developed and has already found various device applications. Properties of SiGe
alloys and Si/SiGe helerostructures were studied in Chapter 4. These heterostructures
can be used for the creation of two-dimensional electron and hole gases and for the
improvement of device parameters.
In Si/SiGe systems, a low-dimensional n-type channel can be obtained if a strained
Si layer is grown upon a strained SiGe layer. Figure 8.14(a) depicts the cross-section
of an n-channel modulation-doped SilSiGe FET. The device is similar to the previously
discussed AlGaAs/GaAs FET with selective doping and improved channel confinement.
Gate
Source [)rain
AuSb -=,
Si
Ii
. Tp1i .
10 nm undoped
'-..=- AuSb
lap layer
Si _l'Gc.. . 100m x=O.5~O graded layer
'
_ Si•.,Ge.~ 10 nm _ Sb doping layer
(a)
Si 20 nm undoped
0.2 lJIll bulTer layer
Si substrate
10' n em
_ _ _ _~in~
r E,
....... 2DEG
---~-~-----'EF
(b) 0.9 eV t
~ 1.1 eV r-E,.
'1d"
Agure 8.14 The Si/SiGe-bascd n-channel MODFET. (3) The cross-section of the device: the
electron channel is fonned in the strained Si layer. (b) Energy-band diagram arlhe device. After
H. Dacmbkes. W. Goodhue et al.• 'The n-channcl SiGc/Si modulation-doped field-effect
transistor,"/EEE Trans. Electron /Jeo,';ces, ED-33. 633--638 (1986). IEEE.
The electron channel is formed in the uodoped strained Si layersituatcd between two SiGe
layers. The lower Sio.7sGe0.2S butTer layer is grown on Si and is strained and uodoped.
The upper Sio.sGeo.s layer is thin and 8-doped by donors. Next, the Sh_xGe.. layer is
graded, with .T varying from 0.5 to O. The top thin Si layer is undopcd. Such a structure
is chosen to create the electron channel and avoid degradation of the characteristics of
the strained layer. Figure 8.14{b) shows the energy-band diagram for this device. The
electron channel on the Sio.sGeoslSi heterointerfacc is markcd. The Sio.sGeo.s donor
layer is 6-dopcd, which leads to the formation of a spike in the potential profile.
There arc many other schemes for Si-based HFETs that combine the advantages of
heterostructure bandgap engineering and sclectivc-doping methods. Motivated by the
8.3 Field·effect transistors 261
AlinAsiGalnAs
'""' MODFET
&!100
o
~
/ AIGaAslGaAs
'-!:::• MODFET
>,
u
c
'"0"'
: :l
10
~
i:l::
o
.!.
::l 10-10
U
1 L-~~~~0..1..-~--->'''''''''''
0.1 1 5
L, La (/lm)
Figure 8.15 The cut-offfrcquency in GHz versus the device length L or the gate length La in ".un
for three groups ofFETs: MOSFETs. MESFETs, and MODFETs. The latter arc represented by
AIGaAslGaAs and AllnAslGalnAs material combinations. The parameters are given at 300 K.
After Fig. 25. S. 1. Peartoll and N. J. Shah, "Heterostructurc ficld-effectlransistors," in S. M. Sze
(Ed.) High-Speed SemiCOlldllCfor Del'ices (New York, Wiley, 1990). Reprinted with pennission
of John Wiley & Sons, Inc.
current wide spread use of Si-based devices, worldwide developments in these areas are
now In progress.
One of the key parameters of contemporary devices is the maximwn frequency of
efficient operation - the cut-off frequency. At fixed device dimensions, the cut-off fre-
quency depends on material characteristics and device design. As a basis for making
comparisons, Fig. 8.15 illustrates representative parameters of all three of these groups
of FETs. One can see that the HFET class exhibits superb high-frequency performance.
Currently, the highest speed is reached for AlInAslGalnAs HFETs with short gates.
This type of HFET with a nanoscale gate of the length LG = 45 nm has a high cut-off
frequency, above 450 GHz.
Nanowire FETs
The previously-analyzed HFETs use heterostructllrcs with two-dimensional electrons.
The next step in reducing the number ofdimensions of the device is to exploit nanowires.
Semiconductor nanowires and carbon nanotubes are attractive components for nanoscale
FETs.
For example, nanowire FETs can be configured by depositing the nanomatcrial onto
an insulating substrate surface, and making source and drain contacts on the ends of the
(aJ
(bJ
Figure 8.16 (a) A schematic diagram of a Si FET with the nanowire, the metal source, and drain
electrodes on the surface of a Si01lSi substmte. (b) A scanning electron micrograph of a Si
nanowire FET; the .scale bar is 500 nm. The 5-nm diamcter Si wire image was obtaincd by using
a high-resolution transmission electTon microscope. Reprinted with penllission, from Y. Cui,
Z. Zhong et lIl., "High performance silicon nanmvire field effect transistors," NlIno Lett., 3,
149-152 (2003). ~American Chemical Society.
nanowire, Figure 8.16(a) illustrates this approach. There, we show a schematic diagram
of a Si-nanowire FET with the mmowire. the melal source, and drain electrodes on the
surface of the Si02/Si substrate, The image oftbe 5-nm-diameter Si wire was obtained
by using a high-resolution transmission electron microscope. In Fig. 8, 16(b), a scarming
electronlllicrograph ofa Si-nanowire FET is shown; the scale bar represents 500 11m. This
approach may serve as the basis for hybrid electronic systems consisting of nanoscale
building blocks integrated with more complex planar silicon circuitry.
An extremely small FET may be built on the basis ofcarbon nanotubes. Indeed, despite
their nanosize diameters, the carbon nanotubes are sufficiently robust and may be long
enough to allow electrical conncctions to lithographically defined metallic electrodes,
therefore making it possible to probe directly the electrical properties orthese nanometer-
size structures. Electrical connections to nanotubes can be achieved either by depositing
a metal electrode over the top of the hlbes (elld-colltacted samples), or by placing the
tubes on the top of prepared metal leads (bulk-colltacted samples).
An advanced nanotube-based FET is shown in Fig. 8, 17. The depicted nanotube FET is
fabricated 011 Si021Si substrates. The silicon is doped heavily to serve as a back gate, The
dielectric Si02 of thickness 10 nm covers most ofthe area of the substrate and electrically
isolates the Si substrate and the wire, Two metal Pd electrodes were deposited to fabricate
the source and drain. This technology provides the nanotube FET with a back gate shown
in Fig. 8.17(a). Additionally, by using atomic layer deposition, an 8-nm-thick H f0 2 film
can be deposited to cover the device from above, Then, 20 11m of Al is deposited onto
the diclcctric (Hf0 2 ) film to creatc the top gate, as shown in Fig. 8.l7(b). Thus, we havc
presented nanowirc FETs with single and double gates. For both devices, the total hlbc
length between metal electrodes was 2 ~1l1, and the top-gated section length was 0.5 11m.
(a) carbon nanotube (c)
S (Pd)
5 (Pd)
(b)
HfO, I
~ckGIlI D (Pd)
Figure 8.17 Schematic diagrams of a carbon-nanolubc FET: (3) Iwo metal (Pd) electrodes arc
deposited onlo a carbon nanotube placed on a IO-nm-thick layer of SiOl that insulates it
from the back gate; (b) 8-run-thick J·lro! serves as an insulator for the lOp gate: and (c) a
scanning electron microgmph orthe carbon nanotubc FET. Reprinted with pcmlissioll, from
A. Jarvcy, J. Guo et al., "Carhon nanotuhe field-effect transistors wilh imegratcd ohmic
contacts and high-k gate dielectrics," NCI/IO Leu" 4, 447-450 (2004). <0 American Chemical
Society.
In Fig. 8. 17(c), an electron-scanning micrograph of a carbon-nanolube FET is shown.

This nanotransistor works and has good performance.
In conclusion, the nallowirc devices discussed here have great potential for applications
in nanoelectronics.
Velocity-modulation transistors
For all of the previously analyzed FETs, the gate voltage controls the !lumber of con·
ducling electrons, and the speed of operation is determined by the transport lime of
electrons along the conducting channel. An alternative principle of device operation
consists of modulating elcctron mobility, or, possibly, electron drift velocity, without
significant changes in the number of current carriers. To achieve such a modulation
of the drift velocity one can usc spatially nonuniform (selective) carrier scattering
across the active channel. Then, a redistribution of the carriers across the channel in
response to a gate voltage would lead to control of the velocity. The time of such a
redistribution across a narrow channel can be considerably smaller than that of the
10ngilUdinai transport. A three·terminal device with modulation of the drift veloc·
ity by an external voltage has been referred to as the velocity.modulation transistor
(VMT).
•
The characteristic time of charge trnnsfer in a channel of thickness about 100 A can
be estimated to be as short as 10- 13 s. Thus, for ultra-high-frequcncy operation, the
velocity-modulation effect should be the dominant factor controlling electric signals.
~
(a)
Gate
~ x
Source z Drain
Barrier
-
-- • -
r
Semi-insulating substrate
I I
I I
I I E N(:xt subband
I I V(unpopulatcd)
(b) I I
I Chi I
--~---- ---~--
~~'~~'
k
Lowest subband
I k I
.1:.1' I .q I
(populatcd) I I
~ PI" ~t;
I I"
(c)
~:
. 'l-- _...:...!._.
P1 --i,:/t2 "
X~O
Agure 8.18 The schematics of QUIT with Iv.'O parnllcl channels: (a) the structure of the device;
(b) electron subbands in different regions of the device and the Fcmli energy (under
equilibrium); (c) an ilIustrntion of the electron paths in the device; IIt'2 and I; ./i indicate the
amplitudes of split and interfering waves, respectively. PI and Pz denote factors of different
phase shifts in the channels.
Quantum-interference transistors
The remaining dass of novel nanoseale FETs presented in Fig. 8.11 is Ihe so·called
qlwlltllm-illleljerellce lrallsislor (QUIT). It is based on quantum ballistic electron trans·
port. According to the classification given in Section 6.2, for such a transport regime,
one has to make the coherence length, '•• greater than the characteristic device scale, L:
I. > L; (8.14)
that is. this device is a mesoscopic device. The principle of operation of a QUIT is the
control o/Ihe inler/erellce pal1ern 0/ conducting electrolls by (III external \'oltage.
Remember that interference is one of the most important effects in the physics of
waves. We studied the interference effect in Sections 2.4. Examples of interference
patterns arc given in Figs. 2.3 and 1.7. In a QUIT, these patterns can be arranged as a
result of interference of the waves traveling through ty,'O or more channels (arms).
A schematic diagram of a n\'O-channel quantum-interference device is shown in
Fig. 8.18(a). Basically, it is an ordinary FET with a short channel and a barrier par-
allel to the current. The barrier is embedded in the middle region of the device. Thc
barrier splits the main channel into two channels: I and 2. There arc also nvo contacts
to the main channel - the source and thc drain. The gate is placed on the top of the
device. Owing 10 the gate, the symmetry of channels I and 2 is broken if a gate voltage
is applied. The distance bct\vcen contacts, L, satisfies the inequality of Eq. (8.14). The
widths of the main channel and the split channels are small, so that there is a quantization
oftrnnsvcrse electron states in the z direction. Two-dimensional subbands are illustrated
in Fig. 8. I 8(b); the subbands, £(kx , k y), and the Fermi energy, E~·, are plotted for three
major device regions. For simplicity, olle can assume that only the lowest subband is
populated in each of the device regions. I f a voltage is applied to the gate, in the mid-
dle region, the bottoms of the subbands, £1 and £2, for channels I and 2 are generally
different. The electron energy spectra are of the form
, ,
£ = E]
tl-(,
+ -- ") =
kx I + f,--" £2
h-(,
+ -- k. 2 + ky') • (8. J 5)
2",· . 2m"' .T
where E is the energy of an incident electron, and k.,. and ky are the components of its
wavevcctor in the plane of the device. Obviously, the energy E does not change along
the channel. The ky-component is also conserved because it is assmned that there are no
forces along the y direction. Thus, from Eq. (8.15), it follows that only the k.t·-components
in the channels can be different.
Let an electron be injected from the source into the left: region of the main channel.
Its wavefunclion is
where XL(Z) is the wavefunction of the lowest subband in the left: region. We can intro-
duce the amplitudes of the waves transmitted into channels I and 2 as II and 12. These
amplitudes determine the electron wave functions inside the middle device region:
upper channel.
lower channel.
where X~l.I(z) and XM.2(Z) are the wavefunctions of the lowest subbands in channels
I and 2, respectively, and XL is the coordinate of the left: end of the channels. Since,
in general, the potential profiles in the two channels are different, the wavcvectors are
different: k:r.l i- k.t.2. Then, let I; and I~ characterize the waves transmitted from the
channels into the right device region; see Fig. 8.18(c). The wavefunction inside the right
region can be written as
(8. J 6)
where PI '= exp(ik.,.1 L). P2 ;: :; : exp(ik.t.2L), and XR(Z) is the wavefunction of the trans-
verse motion at the right end of the channels. For the sake of simplicity, we neglect
multiple reflections from the ends of the barriers.
The wavefunctions XL and XR are normalized to unity; thus, the total transmission
coefficient for the device is
(8. J 7)
We can assume thai both channels, 1 and 2. arc symmetric with respect to z = 0 in the
absence of a gale vohage. Thus, for the lowest occupied subbands we find
and
Then, Eq. (8.17) gives us
IT(E)I' = 2111';1'(1 +e059). (8.18)
where
(8.19)
is the relative phase shift of the two channels. If we introduce the average electron
velocity,
h(k•. 1 + k•. ,)
lJ... =
2m·
we can represcO! the difference between the wavevectors as follows:
(8.20)
ow, the phase shift takes the form

L £1 - £2
9 - -...:....,,.-:.
- v... tI .
The origin of the phase shift is obvious: if £1 f. E2. a difference between the kinetic
energies in the Iwo channels gives rise to different phases of the waves coming into the
right device region; of course, these different phases lead to quantum interference.
For channels that arc symmetric with respect to z = 0, in the absence ofa gate voltage.
we find £1 = £2 = £0, and the phase shift equals zero. If we apply a gate voltage, the
potential energy as a function of the transverse coordinate is modified:
V(=) = Vo(=) - e<l>(=).
Here, <1>(=) is the potential induced by the applied voltage. This leads to subband energies
£1 = £0 - e(X~I.lI<l>lxM.d.
£2 = £0 - e(XI\1.2I<l>IX~1.2).
Using Eqs. (8.10) and (8.19), we find the phase shift,

L e<1>12
9= - . (8.21 )
tI... tl
where <1>12 = (xl\1.2I<l>(=)lx~u) - (x~l.ll<1>(=)lxl\u), The value <1>12 represents the dif-
ference between the average potentials in the channels. This difference detcmlines the
transmission through the devicc.
To calculate the electric current, one can use the results ofScction 6.5, where quantum
ballistic transport was analyzed. Equation (6. t 10) defines the device conductance, G. At
low temperatures, these results lead to the Landauer formula of Eq. (6.121), which, in
the case under consideration, has the form
2e2 4e 2 ,
G= -T(E) = -1"'21-(1 + cosO). (8.22)
It It
The second term in the brackets is due to electron interference. One can see that the
interference controls the device conductance. If the phase shift of split waves () = 0
(<1>12 = 0), the conductance reaches the maximulU, G malt = Sell/ 1f21 2 /1t. If 9 = ]f, i.e.,
trJnJx
e<P 12 = L (8.23)
the interference is destructive and the conductance vanishes.

Let us introduce the characteristic transit time through the channels, ftr = L/v., . If
•
we set L = 2000 A and v-, = 2 x 10 7 cm S-1 , we get Ilr = I ps, and for the destruc~
tive potential difference we get <1>12 ::::::: 2 mY. The shorter channels require a larger gate
voltage. Obviously, the transit time. Itr, determines the cut-off frequency of the device,
Wef = 2IT / f Lr • Since the device channels are not doped and can be made quite short, these
devices portend operation up to the terahertz frequency region.

Our simple model of the quantum-interference transistor allows us to compare this
device with a conventional FET. If the latter is in a normally turned-on stale with the
Fermi energy of electrons in the conducting channel equal to E F , we can estimate the
threshold voltage needed to deplete the FET channel as
fj 2k 2
e<l><h.FET '" EF = F (8.24)
2m-
For quantum devices described by Eq. (8.23), we find
2
7ftw, 1ffj kF 1l 2 k; 2Jr AF
e<Pth.QUIT::::::: eet>12 = L' - m-L - 2m- kFL = ·I-'"ETT·
e<P 1h (8.25)
where AF is the de Broglie wavelength of electrons in the Fermi level. Hence AF « L;

one can see that the quantulU device can operate with a significantly smaller controlling
gate voltage.
Another design of a structure for n quantum-interference transistor is sketched in
Fig. 8.19. The structure is T-shaped and consists of a channel connecting the source
(grounded), drain, and a transverse arm (stub). The transverse arm has finite dimensions
and has a gate on the end of the ann; this end is referred 10 commonly as the "top"
of the device. If the length, Ll' and width, L], are less than the coherence length, 11jI,
the reflection of the electron wave from the arm produces de Broglie wave-interference
patterns. If the width of the main channel, L 1, also is small in comparison with 1IjI,
the pattcrn extends across the channel and determines the transmission coefficient of
electrons through the device, i.e., the source-dmin conductance. A voltagc applied to
the gate changes the penetration length, L-, of the electron wave into the ann and,
as a consequence, the interference pattern. Estimates show that the gatc voltage can
effectively control the conductance of such a T-shaped device when the dimensions L 1,
•
L2, and L] are approximately several hundred Aor less.
L, r-
L)
L,I
-'-
Agure 8.19 The schematics of a T·shaped QUIT. The source (grounded). drain. and gale
electrodes arc indic:ued.
-~ .
~ 1.6
(a) uc
---g
r.
u
1.5
8 q = 0.69 q 0.42
-~.
$ 1.0 ...._........ ················1- ....._....•......
(b) ~ 0.5 :
~ 0.0' .. . . .
-0.28 -0.27 -0.26 -0.25
Gale Voltage (V)
1.8.".------...£'\..--;;::::-;
QD cY 30mK
(e) 1.6 ./')rv~\r~:..11 00 mK
~ ~;;~~~~~~~300mK 500mK
700mK
1.4
-0.50 -0.45 -0.40
Gate Voltage (V)
Agure 8.20 Experimental results on the QUIT effect: (a) and (b) for the two..channcl QUIT T =
0.03 K, (e) for the "T-shapcd" structure, both fabricated from a high-mobility AIGaAslGaAs
heterostructure. (a) Modulation of the device conductance by the gate voltage. (b) The phase
shin as a function or the gate voltage. (e) The same as in (a) for the "T-shapcd" device 31 five
different temperatures. Reprinted with pennission from K. Kobayashi, H. Aikawa et 0/. "Fano
resonance in a quantum wire with a side-coupled quaOlum dot:' Phys. Rev. B, 70. 035319
(2004). (')2004 by the American Physical Society.
As a practical realization of the QUIT effect, we present the results obtained for an
AIGaAs/GaAsstructure. The channels were made from high-mobility GaAs.ln one ofthe
channels a GaAs quantum dot was embedded under the gale. This device corresponds
to the scheme in Fig. 8.18(a). If one of the channels were blocked, one would obtain
the T-shapcd structure, as in Fig. 8.19. Both cases also arc illustrated schematically in
lhe insets of Fig. 8.20. The low-temperature mobility of the Iwo-dimcnsional electron
8.4 Single-electron-transfer devices 269
gas was as high as 9 x lOs em 2 y-Is-I, and the measured electron concentration was
2 x lOll cm- 2 . These parameters correspond to a mean frcc path of about 8 ~m. Low
temperatures (0.03-0.7 K) ensure that the dephasing length I" is even greater. In the
case under consideration, the lengths of the channels are about I ~.un, and the quantum-
dot (two-dimensional) size is 0.2 ~m x 0.2 ~m. In Fig. 8.20(a), the QUIT conductance
is presented as a function of the gate voltage (in units of the quantum conductance,
ell"). One can sec that the conductance is periodically modulated by the gate voltage, in
accordance with the previously developed model of the QUIT. Special experiments with
applied magnetic field, B, make it possible to deduce the phase shift of the electrons in
the QUIT. The results arc presented in Fig. 8.20(b). The modulation of the conductance
for a T-shaped device is shown in Fig. 8.20(c). Measurements at various temperatures
illustrate the smearing out of the interference patterns due to the temperature dispersion
of the energy of the electrons entering the device.
Summarizing, a pronounced QUIT effect is observed for devices with submicrometer
scale at low temperatures. The scaling down of device dimensions will allow significantly
increased working temperatures.
In conclusion, the quantum-interference effect provides a new principle of operation
for three-terminal dcvices. These devices arc now in an early stage of development.
However, these approaches portend effcctive control by a small voltage as well as the
development of very-high-speed transistors.
8.4 Single-electron-transfer devices
Before analyzing examples of the devices based on single-electron transport studied in

Section 6.S, we shall overview briefly a relatively simple technique for the formation of
structures with low-dimcnsional electron gas.
The split-gate technique

Having a two-dimensional electron gas on an interface, or in a quantum well, one can
further lower the electron-gas dimensionality by various methods. One such method
is the so-called spliT-gaTe Technique. The principles of this technique can be explained
as follows. Typically, two-dimensional electrons are separated from the surface of the
sample by a widc-bandgapdielccl'ric layer. It can be a Si02 1ayeron Si, an AIGaAs barrier
layer on GaAs, ctc. Figure 8.21 (a) illustrates such a structure. A modulation-doped barrier
layer decreases electron scattering by the donors and results in high electron mobility. A
thin GaAs layer grown on the top ofthis structure is used as additional electrical isolation
from the metal gates. Let a metal strip- a gate - be deposited onto the top ofthis structure.
The distribution of the potential energy ror the case of a negative applied gate voltage is
presented in Fig. 8.21 (b). According to this energy scheme, two-dimensional electrons
are repelled from the region beneath the metal strip; their Fermi energy, EF. is below
the lowest subband energy. As a result, the region under the gate becomes completely
depleted, as depicted in Fig. 8.21. Now it is clear that, by using several gates, possibly
(a) Thin layer ofGaAs to

z cover AIGaAs
arrow
metal strip
Thin undopcd
d.,
layer of AIGaAs
'--J
(b)
Rgure8.21 (a) Two-dimensional electron gas (2DEG, denoted by dashed lines) at the interface of
a modulation-doped GaAslAlGaAs heterostructure with a depletion region under the negatively
biased narrow metal strip (gate). (b) The band diagnun of the AIGaAslGaAs heterostructure in
the gate region. The lowest subband is shown by the £] line. eJ>b is the built·in Schonky voltage
and eJ>G is the gate voltage.
of different forms, one can create various configurations of regions occupied by the
eJect'rons. It is possible to form wires, dots, rings, cavities, etc. for the two-dimensional
eJectrons. For example, if two closely placed parallel metal strips are fabricated on the
top of the heterostructure, then, by applying negative voltage to these two gates, we can
form two side barriers for the electrons and confine them into a channel. If the channel is
narrow enough, the two-dimensional electrons can be quantized in the second direction
and we obtain a quantum wire; such a wire is shown in Fig. 8.22. The confinement of
the electrons to the dots, wires, rings, ctc. can be accomplished by a heterojunction on
one side and electrostatics on all other sides. The split-gate technique has successfully
been exploited for measurements of transpOri regimes in various quantum structures, for
example, in quantum point contacts and electron waveguides.
A considerably more sophisticated design of patterning of a two-dimensional gas to
a shape desimble for single-electron applications is presented in Fig. 8.23. The main
features of the design shown in Fig. 8.23(a) are the following: (i) the Ohmic contacts
(DC) to the two-dimensional electron gas (contacts to two electron reservoirs (R» and (ii)
a system of gates, which create electrostatic tunnel barriers (TB) and confine electrons
into a quantum dot (QD). The tunnel barriers arc formed when the voltages applied to
the gates are negative with respect to the voltages applied to the contacts. The barriers
should be high enough to decouple the quantum dot and the reservoirs.l.n Fig. 8.23(b), the
8.4 Single-electron-transfer devices 271
Z Quanlumwlrc(IDE
. G)
Metal strips
n ~AIGaAs
Thin layer of GaAs to

cover AIGaAs
Agure 8.22 A quantum wire 011 the imerface of a GaAslAIGaAs heterostructure with two
depiction regions under the negatively biased metal strips.
(a) (b)
1"(2)
£-~ ~ 1:~~ir=;~~~:Gates
'~f:
GaAs Cap Layer
':;:::AIGaAS. n+
Gale
TB
Gale
TB
R R
2DEG in GaAs
---EF
GaAs undopcd
-GaAswarcr
Figure 8.23 (a) Design patterning oftw(}-dimcnsional gas to a shape desirable for single-electron
applications: OCt ohmic contacts; R, reservoir of electrons; QD. quantum dot and TB. tunnel
barrier. (b) The potential profile. V (=), along the line A-B. Ohmic contacts. OC. are not shown
since they are outside of the depicted region.
rcsuhing potential profile is depicted schematically. Split-gate techniques and resulting

structures are used to observe single-electron effects. Indeed, if the quantum-dot-like
structure is small enough. it can be considered to be a single-clectron box. A voltage
applied to the contacts (the source and the drain) induces transfer of electrons through
the structure one by one, as studied in Chapter 6. In general, the flexible split-gate
technique is a powerful method to realize single-electron-transport devices.
Single-electron transistors
Single-electron devices like the one sketched in Fig. 8.23 are two-tenninal devices, i.e.•
they are diode-type devices. It is possible to introduce an additional gate and create
G,
(a)
G,
(b)
Source Drain
Rgure 8.24 (a) The design of a quantum·dot single-electron transistor; (b) an electric circuit with
a quantum-dol single-electron mmsislor.
-c: 12
T 12
~
, Sample 1a Sample Ib
0
~- 8 8
co ""
-""" 4 4
"Uco
0 0 4 12 0 4 8 12
Gate voltage (mV from threshold)
FIgure 8.25 Modulation of conduclance in single-electron tTansislors. Reprinted with permission,
from M. A. Kastner. "The single e1eclron transistor," R('v. Mod. Phys., 64. 849 (1992).
a three-terminal device, a single-electron transistor. In Fig. 8.24(a), the design of the

metallic electrodes on the top of a heterostructure with a two-dimensional electron gas
is shown. The gates G I • Oz, and 0 4 form a quantum dot as above; the gate Gj addition-
ally controls the size and shape of the dot, changing its properties. The single-electron
transistor works as follows.
The electron transfer is determined by two factors: the Coulomb charging ofthe dot and
the quantized energy levels in the dol. If the drain is biased with respect to the source, an
electric current occurs in the regime of single-electron transfer. By applying the voltage
to the gate and changing the quantum-dot paramctcrs, one can change the conditions of
electron tunneling and affect the source-drain currcnt. Examplcs of modulation of the
conductance in single-electron transistors by the gate voltage are presented in Fig. 8.25.
111(: devices have almost the same geometry. Their dimensions are large enough to have
a number of quantized levels. In Fig. 8.25 each peak in the conductance corresponds to
transfer of one electron, when an energy level enters into resonance with the electron
8.4 Single·electron-transfer devices 273
Figure 8.26 A scanning electron microscope image of a single-electron transistor. From Y. Ono,
A. Fujiwara et al., "Manipulation and detection of single electrons for future information
processing," J. Appl. Phys., 97, 0311 01 (2005). Reused with permission from Yukinori Dna,
Akira Fujiwara. Katsuhiko Nishiguchi, Hiroshi Inokawa, and Yasuo Takahashi, JOllI'Ifal of
Applied Physics, 97, 031101 (2005). 102005 American Institute of Physics.
states in the contacts. Though the conductance versus gate-voltage dependences are
different, i.e., not reproducible, the peak spacing is the same for both devices. It is
determined by the change in the gate voltage required to change the charging energy
of the quantum dots by one electron. The figures show clearly that the electric current
is modulated significantly by the gate voltage. Thus, for transistors with single-electron
transport, strong control of very small electric current may be possible. The problem of
fabrication of reproducible devices requires further improvements in technology.
A single-electron pump and turnstile

In the single-electron transistor considered previously, the barriers which separate the
electrodes and the dot practically do not change when a gate voltage is applied (instead,
quantum-dot parameters arc modified). However, variation of the barriers is possible and
it adds a new function to the single-electron devices. As will be shown below, a device
with tunable barrier can work both as a pump and as a (urnstile. Consider such a case for
the example of silicon-based devices. A scanning electron microscope image of such a
device is shown in Fig. 8.26; this single-electron transistor is made in a very thin Si layer.
The conducting channel. source. and drain of the device arc clearly seen in the figure.
The fine gates are made of polycrystalline Si, which is a good conducting material. Three
gates form a dol. Two of them, gates I and 2, tune the barriers between the dot and the
electrodes. Figure 8.27 illustrates the sequence for the pump operation. We start with a
Coulomb-blockade state with an electron in the dot (state 1). Then, by applying a negative
bias to gate 1 and making the left barrier higher, we may close the left channel (state
II). During this action, we keep the dot potential nearly constant by applying a positive
control bias to gate 2. Then, we raise the island potential so that an electron in the dot is
ejected to the right channel, ending up with a new Coulomb-blockade state with 1/ - I
electrons (state III). Next, we open the left channel and close the right channel. keeping
(I)
(II) (V)
( III) (IV)
FIgure 8.27 The sequence for pump operation.
the island potential nearly constant (slate IV), and then lower the island potential so thai
an electron can enter frol11 the left channel (slate V). Thus, during one period only one
electron can be controllably transferred through the device. This system can be viewed
as a single-elecfron pump, since it transfers the electrons one by one without applying a
bias between the source and drain.
The electron pump can have a number ofapplicalions. One ofthcm is a current ctalon.
Indeed, the average currenl flowing through the device is
1= -ef
where J is the frequency of variation of the voltage on the gates. Since the frequency can
be measured with very high accuracy, the device can serve as the elecfr;c·current S((I11-
dartl. The accuracy of such a standard is much higher than the analogous characteristic
of any other current standards.
The tllrJIstile operation is realized by making a small change in Ihe previollslydescribed
regime of operation. The detailed procedure for the turnstile is illustrated schematically
in Fig. 8.28. In the turnstile mode, with a non·zcro source-drain voltage, we first close
gate 2 so that an electron enters only from the left side (stage I). We next close both gates
(stage II), and then open gate 2 so that an electron is emined to the right (stage III). Finally,
we again close both gates (stage IV), and open gate I, reaching the initial state. This
procedure can be accomplished by applying an AC bias to each gate with the phase shift
ofn. Noteworthy is thaI, in this turnstile procedure, at least one of the channels is always
closed, which is not the case in the pump operation mode. This is advantageous because
high resistance effectively prevents the tunneling oftwo c1ectrons- one ofthe major error
sources for single.e!cctron-transfer regimes. Figure 8.29 shows the drain·current versus
drain-voltage characteristics in the aforementioned turnstile mode at several frequencies.
/ = 0.00 I, 0.5, and 1.0 MHz. Staircases quantized in units of ef are observed both for
8.4 Single-electron·transfer devices 275
(I)
-- - - EF
(II) (IV)
EF- - - - -0 ----EF EF- - - - ---- E

F
(Ul)
EF----
EF
Figure 8,28 Schematics of a turnstile operation. From Y. Ono. A. Fujiwara el al., "Manipulation
and detection of single electrons for future information proces~ing;' J Appl. Phys., 97, 031101
(2005). Reused with permission from Yukinori Ono. Akira Fujiwara. Katsuhiko Nishjguchi,
Hiroshi Inokawa, and Yasuo Takahashi, JOlJrJllIl ofApplied Physics, 97, 03110 I (2005). {') 2005
0.4..--~_~_~ _ _-_......,
•••••.••• 1 MHz
0.3 0.5 MHz
~ --- 1kHz ••'
~ 0.2:- ~.;o._-_._y
c 0.1 .. . _
~
~ 0.0 ~ - - -;. -.,.;:-
/ -- --- ----......
'"
u -0.1 ~ - ./.. -
.~ __ i{r__ f
! - - - - - - - - - - - - - -_
~ -0.2 ,:
o -0.3
I
I ,
!
-0.4 ---" '""'~~~~~~~..J
-120 -80 -40 0 40 80 120
Drain voltage (mV)
Figure 8,29 Drain current versus drain voltage characteristics of a turnstile device. 1 pA =
10- 12 A. where pA denotes a picoampere. From Y. Dna, A. Fujiwara er al.• "Manipulation and
detection of single electrons for future infonnation processing:' J Appl. Phys.• 97.03110 I
(2005). Reused with permission from Yukinori Ono, Akira Fujiwara. Katsuhiko Nishiguchi,
Hiroshi Inokawa, and Yasuo Takahashj, JOllrJIal ofApplied Physics, 97. 031101 (2005). «:12005
Potcntial-effect transistors
I~__I_ _-I
Bipolar transistor Hot-electron transistor
I~_I",,--_~
Silicon bipolar Hctcrojunction bipolar
transistor transistor
Figure 8.30 The family tree of potential-effect transistors. The devices are set apan into two large
groups: bipolar and hot-electron transistors.
positive and for negative drain voltages. This is the evidence of single-electron transfer
by the turnstile operation. The levels ofthe current plateaus for the positive drain voltages
are exactly equal to eJ to within the accuracy of the measurement system of about 10- 2 •
In these experiments, the working currents were extremely small.
In conclusion, the main principle of the FETs-current control by means ofa voltage
applied through capacitively coupled electrode(s) - can be successfully applied to single-
electron devices. By combining the field effect and the single-electron phenomena, one
can manipulate the device stales using extremely small electric currents.
To date, single-electron experiments have been conducted at low temperatures (typ-
ically below 20 K). However, with further scaling down of semiconductor technology
to nanosizcs, we may expect that single-electron effects and devices will be exploited at
higher temperatures, including under room-temperature conditions.
8.5 Potential-effect transistors
According to the classification given in Section 8.3, a second large family of electronic
three-temlinal devices is that of potential-elTect transistors. This family is shown in
Fig. 8.30. In contrast to the case of FET devices, potential-effect transistors arc current-
controlled. The controlling electrode is resistively coupled to the active region of the
device and the carriers arc separated by an energy barrier.
The most important representative of this class is the bipolar transistor, which was
invcnted in 1947, and has undergone considerable and persistent transformation. Cur-
rcntly, the bipolar transistor provides high speed of operation in most circuit applications.
Detailed analysis of the physics of bipolar transistors is certainly beyond the scope of
this book. Instead, at the end of this chapter we provide references to books on bipolar
trJnsislors. However, we shall consider the working principles of this transistor, at least
very briefly.
p-n Junctions
The key c1cment of the transistor is the so-called p-lljunctioll. In the simplest case it is
ajunction of two regions in a semiconductor: one side is doped by donors and contains
8.5 Potential-effect transistors 277
Depiction
regions
I•
E<-_ _ -+.....
Electrons
Rgure 8.31 A p-n jUllCtion with abrupt coordinate dc~ndcncc of doping concentration.
(a) (b) (e)

Equilibrium Forward bias Reverse bias
<1>, <1>,
,-....,II f-----, ,-...., 111---
p n p n
E, ----~'::"
\
P \'--
Rgure 8.32 The scheme of a p-njunction (a) al equilibrium; (b) under forward bias ¢I(; and
(e) under reverse bias <1>,_ For non-cquilibrium cases. (b) and (e). instead of onc Fernli level,
there are 1....'0 quasi.Fcnni levels. one for holes, E Fp • and one for electrons, E r... as shown.
electrons: the other side is doped by acceptors and coma ins the holes. In Fig. 8.31, the
scheme of such a p-n junction is presented. An abrupt transition between the p- and
n-doped sides of a semiconductor is assumed. The charge transfer across the junction
lines up the various Fermi levels and results in a Fermi level, E F • It depletes two narrow
regions in the p- and n-doped sides and creates a built-in elcctrostatic potential, <1>81.
The potential barriers of height e<l>BI prevent the penetration of electrons into the p-type
side ofthe structure and of holes into the n-type side of the structure. 0 electric current
passes between p and n regions under such an equilibrium condition; see Fig. 8.31(a).
As is clear from this discussion, the potential profile which appears at the p-n junction
and the possibility of control1ing of it are characteristic for potential-effect devices .
Is.:;:::====1"-----:--
Figure 8.33 The current-voltage characteristic of the p-n junction.
Now, let a voltage be applied to the p-n junction. The corresponding electric circuits
are shown in the upper part of Fig. 8.32. There are two quite different cases. The!orwllrd
billS ofthe p-n junction corresponds to a negative potential, <l> = -r (<l>r > 0), applied
to the n-type side. Potential barriers both for the electrons and for the holes decreasc and
arc equal to e(13l - <Pf). This casc is illustrated by Fig. 8.32(b). A positive potential
(the reverse billS <l>r > 0 of the junction) leads to an increase of the potential barrier
between the nand p parts of the structure, as illust.rated by Fig. 8.32(c). Now the barrier
height equals e( <1>81 + <l>r). The applied voltage is dropped primarily across the depletion
regions because their resistances are much greater than those of neutral nand p regions
of the junction. The voltage breaks down the equilibrium; in particular. there is no longer
a common Fermi level. Instead of the true Fermi level, quasi-Fer"'i levels E Fn and EFp
can be introduced for better understanding of changes in the potential relief, as shown
in Fig. 8.32. As soon as EFn ::f. EFp, an electric current starts 10 flow between the nand
p parts of the structurc. For a forward applied bias, the current increases exponentially
as the potential barrier decreases, and the junction becomes strongly conducting. For
an applied reverse bias, the current is small and saturates at large biases. The current-
voltage characteristic of the p-njunction is presented in Fig. 8.33. The main features of
this characteristic are a rectifying behavior and a strong nonlinearity. This nonlinearity
is used in numerous applications of the p-n diode.
Bipolar transistors
A homostructure bipolar transistor consists of hllO p-n junctions. Usually one of
them - the emitter junction - is forward-biased, while the other - the collector junc-
tion - is reverse-biased. The energy-band diagram for a bipolar transistor is shown
schematically in Fig. 8.34. This diagram clearly illustrates the formation ora potential
for electron and hole motion. The control of this potential is the key element of the
bipolar transistor. The device depicted is of tile n-p-n type, though almost all results are
applicable also to p-n-p devices. This diagram can be easily understood in terms of the
charge-transfer effect between the emitter (len n region), the base (middle p region). and
the collector (right n region). The voltage bias is supposed to be such that an electron
travels from the left n region across the p region to the right n region. An applied forward
Emitter Base Collector
,,,- - -- p
,,
,,,
,
-----_ ..
Figure 8.34 An energy band diagram of a homostructurc n-p-n bipolar transistor along the
direction of electron travel. Quasineutral emitter. base, and collector regions are indicated. The
emitter junction is forward·biased, and the collector junction is slightly reverse·biased.
IC 5
r' 4
3
2
Figure 8.35 The collector current Ie in a bipolar tmnsistor for Ave different base currents, In.
bias to the emitter junction lowers the energy barrier for an electron flowing from the
emitter to the base. Simultaneously, a similar flow of holes from the base to the emitter
appears. Those electrons that overcome the barrier travel across the base. Recombination
of the minority carriers (electrons) occurs in the base, but, if the thickness of the base
region is less than the diffusion length of the minority carriers, these electron losses are
negligible, When the electrons reach the collector junction, they are swept away into the
collector due to the high electric field within the depletion region of the latter junction
and form the collector current. In the bipolar transistor, each of the conducting regions
(the emitter, the base, and the collector) is provided by an electrical contact. Thus, this is
a three-terminal device. The operation of a bipolar transistor is based on the principle of
controlling the current by injecting minority carriers. For example, in an n-p--n device,
the injection of minority carriers from a forward-biased n-p junction (emitter-base) into
the base provides the current through the base contact and the controlling function of the
collector current. In Fig. 8.35, the collector current versus the collector-base voltage,
<PCB, is shown for various currents, Is, through the base contact (compare this with the
current-voltage characteristics of the FETs presented in Fig, 8.10).
The carriers travel from the emitter to the collector perpendicular to junctions; thus.
the carrier t.ransit time through the base determines the speed of operation and the
cut-off frequency of the device. Device scaling in order to reduce the transit time in a
Emitter
Base Base
n Si
11+ Si substrate
Collector
Figure 8.36 The schematic cross-section of an n-p-n SiiSiGe hcterojunclion bipolar transistor.
n++. n+, p+, and n denotc high, intermcdiate (n+ and p+), and relatively low doping,
respectively. The p+ -SiGc layer is strained.
homojunction bipolar transistor can not be achieved while simultaneously realizing all
the requirements for high performance; such requirements include high doping levels
both in the emitter and in the base, and a small base region. The use of heterostructures
helps to solve these problems and to improve device performance. Thus, one can divide
this transistor class into two large and important groups: homojunction and heterojunc-
tion bipolar transistors, as shown in Fig. 8.30. As an example of the latter subclass
of bipolar transistors, consider the Si/SiGe device. In Fig. 8.36, the cross-section of a
Si/SiGe heterostructure n-p-n bipolar transistor is shown. The device is fabricated on
an n+ -Si substrate (subcollcctor with doping of about 2 x 10 19 cm- 3 ) contacted by an
electrode. The collector layer is doped to approximately 2 x 10 16 cm- J and has a width
of about OJ ~n. The 1'+ base is fabricated of a Sil_xGe.1 aHoy with x ranging from
0.2 to OJ, and has a width of typically 30-50 nm. The base doping is in the range from
2 x 10 18 to 5 X 10 19 cm- J • Two-sided metallic electrodes provide the direct contact with
the base region. The emitter is made of n+ -Si with doping of about 5 x JOI7 cm- 3 . A
heavily doped layer, marked as n++ -Si, is placed at the top of Ihe emitter to provide a
good contact to the metallic electrode. Owing 10 the very short base and other pecu-
liarities of the heterojunclions, for such a device an extremely high cut-otT frequency of
75 GHz at room temperature is measured, whereas at liquid-nitrogen temperature, a
cut-off frequency as high as 94 Gl-lz is reported for this type of bipolar transistor.
TIle very recent record results for this type of bipolar transistor were reported by
IBM: the silicon-germanium transistor hits 500 Gl-lz. More detailed analysis of het-
erostrucnLre bipolar transistors can be found in references presented al the end of this
chapter.
Hot-electron transistor;
Another group of potential-effect devices is represented by hOI-eleClrDl1lral1sislOrs, as
illustrated in Fig. 8.30. These devices employ an emitter-barrier-base-barrier-collector
structure. The hot electrons are injected over, or through, an cmjtter barrier into a
(b) (e)
M
s
MOM
o 0 M -- --
Figure 8,37 Energy·band diagrams for metal-base ballistic-injection transistors. Metal, oxide,
and semiconductor layers are marked by M, 0, and S, respectively. Semiconductor layers are
doped. The diagrams arc presented under operational bias conditions.
narrow base region. Injected carriers have a large velocity and transit through the
base almost ballistically; that is, tbey transit the base without significant scattering.
The current through the device is controlled by changing the height of a collector bar-
rier. This principle is used in a number of different configurations and combinations of
materials.
Consider a ballistic-injection device, which is primarily a unipolar device; that is,
only one type of carrier, say electrons, is used. A ballistic-injection device also con-
sists of an emitter, base, and collector. The role of the emitter is to inject electrons
with high velocities into the base; the second electrode should collect these electrons.
The input base voltage controls the electron injection and. therefore, the output emitter-
collector current. Ifsmall changes in the input produce larger changes in the output, the
device exhibits a current gain. Generally, in ballistic devices, the electrons are injected
into the base with a high energy exceeding 0.1 cv. lltis should lead to a decrease in
the timc of flight through the base region. Another advantage of ballistic devices is
related to their unipolar character; this means that it is possible to choose the fastest
type of majority carriers (electrons) and avoid the participation of slower minority
holes.
In order to realize a transistor faster than a bipolar or field-effect transistor, several
schemes of ballistic-injection devices have been proposed. They differ in terms of the
physica I mechanisms of electron injection as well as in the materials used in the devices.
The first and, perhaps, the simplest device is a metal-oxide-melal-oxide-metal het-
erostructure. In Fig. 8.37(a), this device is shO\\11 under a bias. Other similar device
structurcs are the metal-oxide--metal-semiconductor stmcture and the semiconductor-
metal-semiconductor (SMS) structure presented in Figs. 8.37(b) and 8.37(c), respec-
tively. The principle of operation is the same for these three structures. Consider, for
example, the case of the SMS structure. One can see a close analogy with the bipo-
lar transistor: a forward-biased semiconductor-metal junction serves as the emitter, a
second metal-semiconductor junction serves as the collector, and a metal layer is the
base. Bothjunctions are, in fact, Schottky diodes, one is forward-biased and the other
is reverse-biased. Under such bias conditions, electrons are injected over the Schottky
barrier with energies substantially exceeding the thermal energy inlhe base. If the base is
narrow, the electrons "fly" across the base region without losing thcir encrgy. Their sub-
sequent destiny is defined by the base---collector bias: a lowering of the collector barrier
increases the fraction of electrons coming into the collector electrode and, consequently,
the collector current.
Various materials have been lIsed for the structures shown in Fig. 8.37. In particular,
the Si/CoSiz/Si structure was investigated for the SMS scheme. An ex factor as high as 0.6
was reported for this casco A common disadvantage of metal-base transistors is the high
electron reflection from the mctal. This reflection is, mainly, a quantum-mechanical effect
in nahtre and, therefore, can not be avoided. It occurs even for ideal semiconductor-metal
interfaces.
Semiconductor heleroslructures may be used in other ways to realize ballistic-injection
devices. LeI us consider n-type devices. In semiconductor-based ballistic-injection
devices, the emitter, basc, and collector are doped regions separated by two barriers.
The barriers can be formed by growing layers of materials with a positive conduction-
band offset as shown in Fig. 8.38(a). This structure actually uses four heterojunctions.
Another kind of barricr can be produced by planar acceptor doping in a homostructure
as illustrated in Fig. 8.3S(b).
Consider the typical parameters of structures for ballisticMinjection devices. If the
height of the emitter barrier is Vb, the velocity of electrons injected into the base can be
estimated as VB :::::: j2 Vb / m- = 5.9 x 10 7 Vb/(m·1 mo) cm S-1 , where Vb is the barrier
energy in eV. For GaAs we can assume Vb :::::: 0.3 eV and III· = 0.067mo. For the velocity
we gel VB = 1.3 X 108 cms- I • This value is apprcciably larger than the characteristic
elcctron velocity in devices such as FETs and bipolar transistors. Another important
feature is that the injected electrons exhibit velocity spreading in a very narrow velocity
cone. Actually, the average value of the lateral component of the electron momentum in
the emitter is PIJ.E - j2",·kB T. whcre T is the device temperature. Owing to the lateral
translation symmetry this component docs not change under electron injection through
the barrier. Thus, the characteristic angle for the velocity spreading can be evaluated as
follows:
B _ "11.8
vB
= 1'11.8
lit· VB
= PII.E
m*UU
= j k BT .
Vb
For examplc, in the case of liquid-helium temperatures and Vb ~ 0.3 eV, this equation
gives () :::::: 6°.
Collisions in the base reduce the number of ballistic electrons. If the electron mean
free path in the base is Ie and the base width is LB, the fraction of ballistic electrons
collected by the collector is estimated to be a :::::: exp( -Lei Ie). Thus, the base region
should be quite narrow. In this case, the base region has 10 be heavily doped to reduce
the resistance for Ihe base current.
A limitation of the base-doping technique comes about as a result of the fact that a
high doping gives rise to additional electron scattering and quenches the ballistic regime
ofelectron motion. Specifically, for III-V compounds, if impurity concentrations exceed
10 18 cm-J electron scattering bccomes very strong.
As a result, for AIGaAs/GaAs devices. base regions with doping of about 10 18 cm- 3
and with widths oOo-SO nl11 are used. Optimization ofthe structure parameters facilitates
(a)
emitter
- - --
n--GaAs
AIGaAs
n--GaAs IGaA n--GaAs
base
- - - -- collector
(b)
- - --
Depleted
- -n+-- acceptor
layers
- - --
(c)
- -n+--
-
- --
n--GaAs Graded-gap -- -
AI.,Ga,_.,As
GaAs
AIGaAs
• Undoped
Agure 8.38 Energy-band diagrams for semiconductor ballislic tr:lnSiSlors. (a) The AIGaAs/GaAs
structure with fOUf hetcrojunctions and a doped base. Electrons are injected into the base due 10
tunneling. This kind of device is referred to as a tunneling hot-c1cclron·transfer amplifier. (b) A
homostructurc electron device. Barriers are fonned by planar p-doping. The base is n-doped.
(e) A device with an undopcd base. After Fig. 12. S. Luryi, "HoH:!eclron transistors:' in S. M.
Sze (Ed.), High-Speed Semiconductor Devices (New York. Wiley. 1990). Reprinted with
pcnnission of John Wiley & Sons. Inc.
the realization of a ballistic device with a transfer ratio of a = 0.9 at liquid-helium

lemperature.
Figure 8.38(e) presents another design for a ballistic device. The emitter barrier is
made with a graded composition and the base region is undoped. The base is induced
by !he elcctric field of the collector, which leads to the fonnation of a two-dimensional
electron gas at the undoped interface. This type of ballistic device is referred to as the
induced·base frallS;sfor. The advantagesofthe induced-base transistor are the following.
The two-dimensional electron gas in the base is characterized by a high mobility and
large two-dimensional electron concentrations of up 10 2 X 10 12 cm- 2 . Both effects
cause a low base resistance; thus, the base can be choscn 10 be very narrow. typically
•
100 A. This results in a high fraction of ballistic electrons coming into the collector.
For example, a well-designed AIGaAs/GaAs induced-base transistor results in Ct ::::: 0.96
even at room temperature. Similar induced-base transistors have been fabricated using
InGaAs/AIGaAs and p-doped Ge/SiGe heterostructures.
In conclusion, hot-electron transistors with nanoscale base regions are characterized
by a high operation speed. An electron transit time ofthe order of 0.1-0.5 ps corresponds
to a cut-off frequency in the range 2-10 THz.
8.6 light-emitting diodes and lasers
So far we have studied elecmmic nanoscale devices, i.e., a class of devices that
exploits electrical properties of nanoslructures and operates with electric input and out-
put signals. Another class is composed of optoelectronic devices, which are based on
both electrical and optical properties of materials and work with optical and electric
signals.
In this section, we will analyze two very important classes of optoelectronic devices:
light-emitting diodes and laser diodes. As their titles imply, the devices were invented
to produce light with certain properties. In particular, the energy of the electric
current flowing through these diodes is transformed into light energy. These opto-
electronic devices have a huge number of applications and deserve consideration in
detai I.
In Chapters 2 and 3, among various different wave fields, we inlroduced and studied
electromagnetic fields, of which light waves arc a particular example. In addition, we
studied e1cctrons in atoms (Chapters 2 and 3) and solids (Chapter 4). However, it was
supposed that these two physical entities do not interact with each other. In fact,light-
maner interaction gives rise to flmdamental physical phenomena. Historically, light-
maner interactions provided some of the first evidence for the quantum nature of maner.
Remember that an electromagnetic field consists of an infinite number of modes (waves),
each of which is characterized by a wavevector and a specific polarization. According to
quantum physics (see Chapter 3), each mode may be described in terms of a harmonic
oscillator of frequency w. Correspondingly, the energy separation between levels of
this quantum-mechanical oscillator is flw; see Eq. (2.41). This oscillator may be in
the non-excited state. which manifests the ground-state or zero·poi,,' vibrarions of the
electromagnetic field. The oscillator may be excited to some higher energy level. If the
Nth level of the oscillator is excited, there are N quanta (photons) in the mode under
consideration. The classical description of an electromagnetic wave is valid at large
numbers of photons: N » I.
Besides quantization of the energy, the quantum nature of electromagnetic fields is
revealed in the equilibrium statistics of photons. Indeed, photons obey the so-called
8.6 Light-emitting diodes and lasers 285
E,----.....- - -
E, 1- _
Figure 8.39 A two-level system interacting with light.
boson statistics. or Bose-Einstein statistics, which gives the average number of photons
of some chosen mode under equilibriulll in the fOfm
I
N - ----;:----::-:-=-----, (8.26)
- exp[t",,/(k B T)J - I'
with T being the ambient temperature. Equation (8.26) is also known as Planck 'sJormllla.
Clearly, the Bose-Einstein statistics of photons differs drastically from Fermi statistics.
which we used for electrons (sec Chapter 6). In particular, it permits the accumulation
of an arbitrary number of photons in any mode. For example, from Eq. (8.26) it follows
that at kB T j(tIClJ) » I the phonon number N ~ kBT j(ttw)>> I, whereas the number of
electrons occupying any state can not exceed I.
After these introductory remarks on quantum properties of electromagnetic fields, we
review briefly the three major processes involving a quantum material system and light:
absorption, spontaneous emission, and stimulated emission.
Photon absorption and emission

In order to visualize these processes, we considcr a simple quantized system with two
energy levels £1 and E 2 as depicted in Fig. 8.39. TIle occupancies of the energy levels
of this system correspond to particular states of a system of the electrons. The charged
elcctrons intcract with thc electromagnetic field. This interaction results in transitions
between quantum states of the system. These transitions arc frequently referred to as
phototra"sitio/ls. According to the quantum theory, the system can change ils energy as
a result of interaction with electromagnetic waves exclusively of the frequency
w = (E, - E,)/fl. (8.27)
If the lowesl energy level £1 is occupied, the wave can excite the systcm into the upper
level £2 and the electromagnetic energy must decrease. One can describe this process as
the absorption ojolle photo" because the energy of thc electromagnetic field decreases
by E 2 - E1.lfthe system occupies the upper level £2, it can make a transition to level
£ I as a result of interaction with the electromagnetic field. Then, the electromagnetic
energy increases by £-2 - £1. This process represents Ihe em;ssio" oj a photoll with
energy tlW. When activated by an external electromagnetic wave, the latter process is
called stimulated emission. It is important that, for stimulated emission, each emitted
photon has the energy, direction, polarization, and even phase coinciding precisely with
those of the stimulating wave.
Both processes, absorption and stimulated emission, can be described in terms of

an interaction with a classical electromagnetic wave. The rates of these proccsses are
proportional to the intensity of the wave, or, according to Table 2.1, proportional to the
!lumber of photons Nq.b of a givcn wavcvector q and polarization b. These rates can be
written as
RatH. = BI2Nq.bIl1. (8.28)

R SI. rnl = B ZI Nq.bIl Z, (8.29)
where III and 112 are the numbers of particles in the system occupying the levels I and 2,
respectively; B 12 and BZ I are kinetic coefficients describing these processes. The physical
meanings of these coefficients will be addressed in the subsequent discussion.
The two processes of absorption and emission are insufficient to describe the whole
picture ofthe interaction betwcen I1ldiation and matter. For example, Ict us apply these two
processcs only for thermal equilibrium conditions, for which the ratio of the populations
of the two levels is
112 = e-(E!-E,)/(kIlT) = e-flw/(kBT)
(8.30)
III
(see Section 6.3 and particularly Eq. (6.14». Using Eqs. (8.28}-(8.30) onecan sec that
Rabs =F RSl. em at any tcmperaturc T. This rcsult is in contradiction with the expected
equilibrium bctwecn the system and the field. According to the Einstein theory, there is
an additional quantum radiative transition in the system with the spontaneous emissioll
ofa pltoton of the same mode. The rate of this process is
R sp .em = A21 11 2. (8.31)
where A21 is the coefficient or rate of spontaneous emission. The spontaneous process
does not depend on the intensity of the electromagnetic wave and takes place even in
the absence of this wave. According to quantum electrodynamics, the excited matcrial
system spontancously cmits a photon as a result of the interaction between the electrons
and the zero-point vibrations of electromagnetic fields. The zero-point vibrJtions of
electromagnetic fields were introduced briefly in Section 3.3.
In contrast to the case of stimulated emission, a photon produced by the spontaneous
process has an arbitrary phase. Moreover, this process produces photons with different
directions of q and polarizations, but the energy is fixed, i.e., it produces photons of
different modes of the same frequency.
Now we can apply the results of Eqs. (8.28)-(8.31) to thermal equilibrium. Under
equilibrium conditions, the total rate of photon emission has to be equal 10 the rate of
photon absorption; thus we will have
Rabs = Rsp.em + Rst.em· (8.32)
Using the Planck formula of Eq. (8.26) and the ratio of liz! 111 of Eq. (8.30), and sub-
stituting the expressions for Rabs. Rsp.enh and RSl.rnl into Eq. (8.32). one can find thc
relation
(8.33)
Because this relation has to be satisfied at any arbitrary temperature T, one can obtain
two equalities:
(8.34)
Thus, we have established the existence of three basic processes for the resonant interne·
tion of radiation and maner: absorption, stimulated emission, and spontaneous emission.
Moreover, we have found the relations between the coefficients determining the rales of
these processes. It is worth emphasizing that all three processes are related to interactions
with photons of the same mode.
The sum of the stimulated and spontaneous emission rates, as determined by
Eqs. (8.29) and (8.31), gives the total emission ratc of photons for a fixed mode:
(835)
From this equation. one can see thai stimulated emission dominates over spontaneous
emission for a fixed mode, {q. bl. if the number of photons, N q.b , is sufficiently larger
than I. However, there is spontaneous emission of a great number of other modes with
the same frequency but with different directions ofq and different polarizations b. This
total spontaneous emission can be the dominant radiative process even if stimulated
emission is the most important process for a particular mode.
Now, let us compare absorption and stimulated emission by calculating the rate of
increase of the number of photons in some fixed modes:
(d:;.b )Sl = R = Rsl.tm - Rabs = 821 Nq . b(n2 - Ill)' (8.36)
This result shows that. if
Jl2-JlI>O. (837)
stimulated emission dominates over absorption. Evidently, under equilibrium conditions,

the opposite is true.
The inequality of Eq. (8.37) is the criterion for population inversioJl. If a popula-
tion inversion is achieved, electromagnetic waves with the resonance frequency can be
amplified when passing through the material medium. This process of amplification of
the radiation due to population inversion is the key mechanism underlying the operation
of a laser (light amplification by stimulated emission of radiation). The medium where
the population inversion occurs is often called the active medium.
Actually, Eq. (8.36) describes an increase/decrease in the photon number with
time. Amplification/absorption of light waves can be naturally described as an
increase/decrease of their intensity with spatial coordinate along the direction of their
propagation. The relation between the photon numbers and the intensities of the light
waves was introduced in Section 2.4 (see also Table 2.1). The corresponding equation
for the intensity can be derived easily from Eq. (8.36), if we consider a light pulse prop-
agating through the medimll. Lei the maximum of the photon number for this pulse be
characterized by the coordinate x(t). Then, in Eq. (8.36) the derivative with respect to
time can be calculated as

d dx d d
- = - - =Cm - ·
dl dl d¥ dx
where Cm is the velocity of light in the medium. Instead of the photon number Nq.b , we
introduce the light intensity 'I = cmtlWNq,b/V, where V is the volume of the medium.
Now we can rewrite Eq. (8.36) in the form
elI
- = aT. (8.38)
d¥
Here,
I
a = -B2I N q.b (1l2 -"d
em
represents the gain coefficient. The solution ofEq. (8.38) is
'I = 'Ioea .,.
with 'Io being the intensity at a point x = O. Thus, under population inversion the gain
coefficient, a I determines the exponential increase of the light intensity. If a < 0, then,
instead of amplification. absorption of light occurs.
A medium with population inversion can be used in two ways. First, an active medium
can amplify an external light beam in accordance with the previously discussed depen-
dence of'I(x). Second, an active medium can generate a light beam itself, if the proper
optical feedback is provided. Here, the optical-feedback phenomenon implies that some
portion of the energy of the light which is amplified in the active medium is returned into
the medium for further amplification. In an active medium with optical feedback, spon-
taneous emission gives birth to the initial photons. These "seed" photons arc amplified
due to the stimulated emission. and then they partially return to be amplified again. As
a whole, this process leads to the generation of coherent laser light.
Typically, optical feedback is realized by placing the active medium in an optical
resonator. In the simplest case, the resonator consists of plane or curved mirrors, which
provide repeated reflections, and some kind or"trap" of the light - a covity- in the region
between the mirrors. The optical waves which can be trapped in this cavity compose the
resonator modes. A universal characteristic of a resonator mode is the qualityfactor, Q,
which can be defined as
field encrgy stored in thc cavity
Q= w x ... . (8.39)
power diSSipated In the resonator
Dissipation of electromagnetic energy is caused by many factors: absorption by the
mirrors or by mailer inside the cavity, transmission of light through the mirrors, light
scattering, radiation out of the resonator as a resull of the diffraction of light, etc. The
power dissipated and the quality factor Illay be different for different modes.
In principle, in a resonator there is a large number of modes with different frequencics
and polarizations, and all possible propagation directions. Excitation of many modes
would lead to cxtremely incoherent emitted light. To avoid such an effect, one can
employ so·caIled open optical resonators. The simplest open resonator consists of two
(aJ (bJ
= f:: - :
- I- - :
- I- -
- I- -
=• t =•
• •
Lz
Figure 8.40 Longitudinal and transverse modes in the planar Fabry-Perot resonator.
(a) Longitudinal modes. Strictly perpendicular light rays do not escape from the resonator.
(b) Transverse modes. Slightly inclined light rays eventually escape from the resonator and have
poor quality factors.
plane mirrors parallel to each other, whjch arc finite in their transverse dimensions; an
example is given by the so-called Fabry-Perot eta Ion shown in Fig. 8.40. In Ihis resonator.
1110st of the modes propagating through the cavity are lost in a single traversal of the
cavity since the mirrors are inclined with respect to the mode-propagation directions. This
implies that most ofthe modes, so-called transverse modes, have a very low quality factor.
Only the waves propagating perpendicular to the mirrors can be reflected back and travel
from one mirror to another without escaping from the resonator. These waves correspond
to the so-called longitudinal modes ofthe resonator. Thus, for finite dimensions ofmirrors
only longitudinal modes can have a high quality factor. Their loss and diffraction are
caused by absorption by the mirrors, by transmission through the mirror, and by wave
diffraction on the sides of the mirrors. The wave·diffraction losses can be made much
smaller than can those arising from other loss mechanisms. 11ms, the open resonator
provides for strong discrimination between modes. Relatively few of these modes have
a high quality factor, Q. According to the definition of Eq. (8.39), they are capable of
accumulating the light energy. The photons corresponding to these high-quality modes
can be generated effectively by the active medium.
At Ihis point, we end our discussion of the simple two-level model of the optical
medium and start to consider more realistic systems.
Intemand emission and absorption in semiconductors

Semiconductors constitute a material system that may be used in practice to realize
controllable light emission. Thus, we shall consider the mechanisms of absorption and
emission of photons in semiconductors. Among these mechanisms, the most important
is interband (bal1ll.to-band) p!totolftlllSitiolls. Absorption of a photon can result in the
creation of an electron in the conduction band and a hole in the valence band, i.e., an
electron-hole pair. The inverse process is radiative electron-hole recombination, which
results in the emission of a photon (sec Fig. 8.41).
Figure 8.42 illustrates the dependence of the absorption coefficient on photon energy
and wavelength for various semiconductors. From Fig. 8.42, one can see that the
8 8
Emission
'II\rII\I'-.
FIgure 8.41 Absorption and emission of photons in a semiconductor: band-Io-band transitions.
absorption increases sharply in the short-wavelength region. Let £g be the bandgap ofthe
semiconductor. Then the material is relatively transparent for liw < Ell,' For t,w > E g,
the semiconductor exhibits strong absorption; accordingly, tlwg = E g corresponds to
the abSOl7Jtioll edge. The shape of the absorption edge depends significantly on Ihe
structure of the electron bands. Remember that in Section 4.4 we defined dirccl- and
indirect-bandgap semiconductors. Direct-bandgap semiconductors such as GaAs have
a more abrupt absorption edge and a larger absorption value than do indirect-bandgap
materials, of which Si provides an example. We can introduce the so-call cd bandgap
wavelength, or cllt-offIl'Gvelength,).,g = 2rrcftjE g. If E g is given in eV, the bandgap
wavelength in micrometers is
A,, __ 1.24 (8.40)

E.
The values of E g and Ag for various III-V semiconductor materials are apparent from the
curves plotted in Fig. 8.42. One can see that intcrband transitions in III-V compounds
cover a wide range from infrared to visible spectra. Optical activity in this spectral region
is crucial for optoelectronic applications of these materials.
A photon absorbed during an interband transition excites an electron from the
valence band to the conduction band, i.e., it creates an electron-hole pair as depicted in
Fig. 8.43(a). The inverse process - the phototransition ofan electron from the conduction
band to the valence band - is referred to as the Ituliative recombination (annihilation)
of an electron and a hole; this is depicted in Figs. 8.43(b) and 8.43(c). According to the
general properties of phototransitions studied at the begining of Section 8.6, there exist
two such proccsses: spontaneous and stimulated emission, as illustratcd by cases (b) and
(c), respcctively, in Fig. 8.43.
From the above analysis it is clear that, to obtain intense interband cmission, one
should provide nonequilibriwn electrons and holes. Moreover, large concentrations of
Wavelength (11m)
64 3 2 1.5 1.0 0.8 0.7 0.6 0.5
~
,
~
-
<::
u
10'
-
~
<::
u
u
10'
InP
Eu
0
10'
u
<::
.--0
t:- 5i
0
~
InAs Ge GaAs GaP
.<>
< 10
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
Photon energy (e V)
Agore 8.U Absorption coefficient versus photon cnc'l.'Y and wavelength for intcrband
photomlnsitions in various semiconductors. After G. E. Stillman, V. Robbins el al., "'III-V
compound semiconductor dcvicc=s: optical detectors:' IEEE Trolls. Electroll Devices. ED-31.
1643-1655 (1984). IEEE.
E (a) (b) (c)
E, -~
E,
E,
~~[ ~
E,
0 o o
k k k
Agure 8.43 (a) The absorption of a photon rcsulls in c~citalion of the eleclron-hole pair.
(b) Spontaneous emission, which can be interpreted as electron-hole recombination.
(c) Stimulated emission.
both electrons and holes in the same spatial region correspond to the population inversion
necessary for stimulated emission. Indeed according to Section 4.4. the presence of the
holes means "empty" electron states in the valence band. This explains the appearance
of the population inversion between the conduction and valence bands under conditions
of large electron and hole concentrations.
However, noncquilibrium carriers quickly relax due to various interband relaxation
mechanisms. \Vhen the electron and hole concentrations are highly nonequilibrium. the
characteristic lifetime of these excess carriers is small. To describe the noncquilibrium
electrons and holes. we introduce the density of the pumping (excitation) rate, 'Rpwnp,
which represents the number of electron-hole pairs excited in a unit volume per unit
time. For the case when the radiation mechanism is the dcc<ly of electron-hole pairs, one
can write
(8.4 J)
where II is the concentration ofelectron-hole pairs, B is a parameter, and Til. = 1/(8n) is

the radiation lifetime. For the concentrations of about 10 18 cm- J which arc encountered
in practice and are necessary to realize the needed significant pOplLlation inversion and
to achieve lasing conditions, this time is less than 10 os. The pumping rate, 'Rpump •
necessary to induce such a concentration is estimated to he
(8.42)
To obtain an idea of how intense this pumping is, we can perform the following esti-
mates. Suppose that the bandgap, £g, is abollt I c'Y, then the pumping rate of Eq. (8.42)
corresponding to the density of the pumping power given in the previous example is
Eg'Rpump ? 16 MW cm- J • This is a huge pumping power!
At Ihis point, an important spectral property of light amplification in semiconductor
materials will be discussed. For the previously analyzed two-level model, phototran-
sitions and light amplification/absorption were possible only at a fixed photon energy
f/w = E 2 - E]. This implies that in such a two-level system the amplification and absorp-
tion occur in a very narrow spectral range nearlhe value (£2 - El)/ft. On the contrary, in
semiconductors, the spectral range within which amplification and absorption are possi-
ble is restricted only from below, w > Eg/ft. Thjs results in a wide spectral band ofoptical
activity. In Fig. 8.44, we show the spectral dependences of the gain/loss coefficient for
GaAs for various electron-hole concentrations at room temperarure. The left boundary
ofthese dependences coincides approximately with the bandgap value 1.42 eV. The spec-
tral range for which the gain coefficient, Ct, is positive becomes wider and its maximum
moves to larger photon energies as the electron/hole concentration increases because the
number of inversely populated states increases with the concentration. Amplification in a
wide spectral range allows one to build a laser with considerable tuning of the frequency.
The following two methods can be applied to generate large nonequilibrium carrier
concentrations. The first method is optical excitation or pumping, whereby external light,
often incoherent, with a photon energy larger than the bandgap is absorbed and creates
nonequilibrium electron-hole pairs. Then, because of the short time associatcd with the
intraband relaxation, the electrons and holes relax to the bottom of the conduction band
and the top of the valence band, respectively, where they are accumulated. If the rate of
optical pumping is sufficiently large, the neccssary levcl of intcrband emission can be
reached. In particular, invcrsion between the bands can be induced and light amplification
and laser generation become possible. The optical pumping corresponds to a conversion
of onc kind of radiation, not nccessarily coherent, into coherent radiation with a lower
photon frequency. One employs optical pumping in cases when electric-current pumping
either is not possible or is ineffective. This pumping method is often used to test prototype
laser structures before the dcsign of thc current pumping system.
160,--,..-,...-,...-,
120 ,..,
'J'
7-
80 ...-o~
oF
40
%
I
~
"
u
~
'"
~ 0 -- -- -- -- -
40
% \
\
I \
-80
1.42
'1.46
"
,
\ \ 1.50
Photon Energy (e V)
Figure 8.44 The amplification coefficient versus the photon energy in a GaAs light amplifier for
five different carrier concentrations. The temperature is 300 K. The results are given for five
values of the concentration with equal steps orO.25 x 10 18 cm- 3 • After 1. Singh, SemicOllllllcror
Devices: An ImroduClion (New York, McGraw-Hili, 1994), Fig. 11.16.
laser diodes
A much more convenient method that can be applied to achieve the interband emission is
electron and hole injection in devices with p-njullclions. The physics ofa p-njunction
was studied in the previous section. There, we found that, under forward bias of a p-n
junction, it is possible to injeclthe electrons from the n part of the junction and the holes
from the p part of the junction into the same spatial region. Such a double injection can
give rise to large nonequilibrium electron and hole concentrations. Often, the nand p
parts arc separated by a narrow intrinsic (undoped) i region. The corresponding stmcture
is called a p-i-n device.
In Fig. 8.45, we show the case of forward biasing of a p-i-n structure - the so·called
jlar-band condition, when there are no potential barriers for the electrons and holes.
and the maximum possible carrier injection into the i region is realized. In the case of
direct-bandgap semiconductors, double injection provides for the intense emission of
light.
It is easy to estimate the electric currents which arc necessary to obtain a given
concentration of carriers under double injection. We define the active region, where both
noneqllilibrililTI electrons and holes are present. In the case of a p-i-n structure, the
active region coincides with the i region. Let the length of this region be I. We introduce
E, \.
•
\.
e
p+ .
I n+
E, @ ,
~
= - \.
Figure 8.45 The double injection of the electrons and holes, which leads to the formation of
intcrband population inversion in an active region.
a cross~sectional area A, through which the current I is injected into the p-n region of a
diode. Then, at steady state, the rate of injection of electrons and holes into a unit volume
per unit time can be expressed as
I J
R---- (8.43)
-eAI- el'
where J = 1/ A is the injection current density. Injection leads 10 the accumulation of

nonequilibrium electron-hole pairs with concentration
(8.44)
where r is the tOlallifetime of the nonequilibriul11 pairs in the active region. In general,
due 10 the small electron and hole lifetimes, the active region is quite narrow. For GaAs,
the length of the active region is estimated to be frOIll I 1-.1111 to 3 I-llll, depending 011
temperature, crystal quality. etc. Now, we can usc Eq. (8.43) to estimate thc current
density necessary to achieve the pumping rate ofEq. (8.42): J ~ (1.6-4.8) x 103 A cm- 2 .
The current densities obtained are very large. These estimates indicate that, in order to
operate with acceptable current levels, the cross-sections of real emitting diodes have to
be very small.
From Eqs. (8.43) and (8.44), it follows that another possible way of decreasing the
pumping electric current is to reduce the thickness ofthe active region I. This thickness is
one of the critical parameters for injection pumping ofJight-emitting devices. Indeed, as
understood from the previous discussion and Fig. 8.45, an excess carrier concentration
in the active region always leads to carrier diffusion out of thjs region: the electrons
diffuse through the active region to the p part of device, while the holes djffuse to the n
part. In general, "di ffusive leakage" of the carriers is limited by their finite lifetime r(II).
In Section 6.2 we found that, if the diffusion coefficient 0 is given, the diffusion length
during a time r can be estimated as L D = Ji5T. In the case under consideration, T is the
excess carrier lifetime and the length Lp corresponds to the average distance of electron
(hole) transfer before recombination. Obviously, the minority carriers (the electrons in
the p part and the holes in the n part) recombine immediately with the majority carriers
(the holes in the p part and the electrons in the n part). Thus. the width of the region with
the excess carrier concentration can not be less than the diffusion length L n, i.e., d > Ln.
Since the diffusion length of electrons and holes in direct-bandgap materials is of the
order of a few micrometers, it is not possible to make the active region shorter than a
E
Barrier
- - -- -----_ .. -
liw ~ol
._._-.- . . . . __ .
--_ .. _--_.-
r
-
Figure 8.46 The cnergy·band diagram of a double heterostructure for light amplification!
generation. The applied voltage c1>o induces the flat band conditions; Eill. £,2. and Ell arc the
bandgaps in different regions orthe structure. The energy states filled by the electrons arc
denoted by the shaded areas. TIllis. in some spatial region there arc electrons in the conduction
band and depopulated siaies (holes) in the valence band.
few micrometers. This conclusion is valid for homostructure (bulk-like) semiconductors.

Heterostructure technologies open the way to different solutions 10 this problem.
In order to localize nonequilibrium electrons and holes in a smaller active region, one
can employ two hctcrojunctions. The basic idea of using a double heterostructure is to
design potential barriers on both sides of the p-n junction - this prevcnts electrons and
holes from diffusing. The potential profile of a double heterostructure is sketched in
Fig. 8.46. The heterostructure consists of three materials with bandgaps £gl, Egl, and
E 8 ]. The band offsets are chosen appropriately to design a structure with a barrier for
electrons in the left part of the structure before the p region and a barrier for holes in the
right part before the n region. The middle i region with the bandgap £&2 is accessible
by both lypes of carriers and it serves as the active region. Figure 8.46 corresponds
to the case of the flat·band condition of a p-n junction with a double heterostructure
embedded in the depletion region. In this case, it is not the diffusion length but the
distance between the barriers that determines the size of the active region. As a result.
the size can be as small as 0.1 ~m and the critical elcctric current is smaller by one
order of magnitude or more compared with that of a conventional homostructure p-n
junction.
Ifthe thickness ofthe double heterostructure is decreased further, the influence ofquan·
tum effects on carrier motion becomes important. For the structure shown in Fig. 8.46,
quantum effects do not lead to any advantages because the advantages can be obtained if
heterostructures are designed so that quantum confinement applies both to electrons and
to holes. The simplest case of quantum confinement can be achieved if a quantum-well
layer is embedded in an active region of a Iype-I heterostructure. Three possible designs
of active regions exhibiting quantum confinement are sketched in Fig. 8.47. Cases (a)
and (b) correspond to the resultant confinement in a single quantum well, while case
(c) corresponds to confinement in multiple quantum wells. For these designs, electrons
t
0. -
~
(a)
r
(b)
-
(e)
r
e
- :;:
'"
~
0,
L
-<
~
.-<:'<
Agure 8.47 Composition profiles of(AI~ GaI-.. )o.slnojiP h~leroSiruclUrcs providing
simullaneously quantum confinement orlhe carriers and optical confinement: (a) a single
quantum "'t:1I and a step-like refractive-index heterostructure. (b) a single quantum well and a
grndcd-indcx optical confinement structure, and (e) a mulliple quantum well and a step-like
index structure. After P. S. Zory Jr.. Quantum Well Lasers (Boston. MA. Academic. 1993).
SQW L.-50A
200
~
, InGaAsP GaAs
E A - 1.3 Jlm
~
u !OO
c
!:!
0
0 20 40 60
i (A el11-')
Figure 8.48 The amplification coefficient «0 as a function of the pumping current density for
•
single InGaAsP and GaAs quantum wells of width 50 A. After P. S. Zory Jr.. Qllall1ll1/1 Well
Lase,:~ (Boston, MA. Academic. 1993).
and holes thaI are either generated by external light or injected from p and n regions
move in barrier layers and then are capturcd in the active region and quantum wells. The
characteristic time of this capture is less than I ps. Escape processes require additional
energy and have low relative probabilities of occurrence. Carriers in the quantum wells
relax to the lowest energy states available. This results in the accumulation of both types
•
of carners in an extremely narrow active region. which is typically 100 A wide or even
narrower. A similar situation can be realized if quantum wires or dots are embedded in
the active region.
A positive effect of electron confinement in quantum wells is illustrated in Fig. 8.48.
In this figure, thc results for the gain factor ao as a function of the current density J
are presented for two material systems: InGaAsP/lnP and AIGaAslGaAs heterostruc-
tures. It can be seen that the gain coefficient always increases with the current. The
current for which ao becomes positive. i.e., a population inversion is cstablished, can
be defined as the threshold current. The gain coefficient is shown for hct'crostruc-
•
tures with 50-A quantum wells. This system has low values of threshold currents for
both heterostructures: 20 A cm- 2 and 60 A cm- 2 for InGaA InP and AIGaAslGaAs.
Cleaved
surface
d
+
L, P n
, ,
J , ,
Cleaved
surface
Agure 8,49 The scheme of an injection laser with lWO cleaved facets that act as reAectors. After
B. E. Saleh and M. C. Teich, Flllldamellfais a/Photollics (New York, Wiley. 1991). Reprinted
with permission of John Wiley & Sons, Inc.
respectively. For both material systems, the ao(J) curves increase sharply with the cur-
rent density above the threshold. As discussed previously, lasing can be achieved if a
light amplifier is supplied with a path for optical feedback. For an injection laser, the
feedback is usually obtained by cleaving lhe crystal planes normal to the plane of the
p-n junction. Figure 8.49 depicts a device with two cleaved surfaces forming an optical
resonator. For the light reflected from the crystal boundaries, we define the reflection
coefficient by
,. -- 7,
--
I -
in
where ~n and It are the intensities of incident and reflected light, respectively. The
reflection coefficient for an ajr-semiconductor boundary is
r= ("" +_1)'
"ri I
where nn is the refrdctive index of the semiconductor material. Since semiconductors
usually have large refractive indexes the coefficients r are large enough. The intensity of
the liglll transmitted through this mirror is
Let two cleaved surfaces be characterized by two coefficients, rl and r1. After two passes
through the device, the light intensity is attenuated by the factor"l x 1'2_ We can define
an effective overall distributed coefficient of optical losses:
I1 r 1 (I)
= --In
2L.,
-- .
1'11'2
Here, Lx is the distance between the cleaved surfaces. In principle, there can be other
sources of optical losses in the resonator. Let them be characterized by the absorption
3r------------,
:;:
5~
2
"0~
(a)
-&"
0.
1
0"
o 2 4 6 8 10 12
Injection current (mA)
Figure 8.50 The laser output versus the pumping current for a GaAs single-quantum-welliaser.
After B. E. Saleh and M. C. Teich. FlIIullllllc>lItllls ofPhotonics (New York, Wiley, 1991).
Reprinted with permission of John Wiley & Sons, Inc.
coefficient, as. Then, the total loss coefficient is
If a is the gain coefficient of some light mode in this resonator, we can write the criterion
jor/llser oscillatiolls as
a :::: alhreshold = ar + as· (8.45)
For injection lasers, the criterion of Eq. (8.45) is a condition imposed on the magnitude
of the injection current density, J.
In order 10 have an idea of the order of magnitude of optical losses in laser diodes, we
consider an example. For light of visible and ncar-infrared ranges, the typical refractive
index can be estimated as lin::::: 3.2-3.5. Thus, the reflection coefficient isr :::::: 0.3. Let a
diode have the geometry presented in Fig. 8.49 with dimensions such that d x Lx x L y =
I ~m x 200 !-till x200 J.1m. Then, we can estimate the radiation losses: a r ;:::: 60 cm- I .
TIle total current through the laser diode is 1-2 A. The threshold currents for lasing
for the quantum-well-based structures are considerably smaller and are typically of the
order of tens ofmA. In Fig. 8.50, the output light power is presented as a function of the
•
injection current for an AIGaAs laser with a single IOO-A quantum well embedded in
the active region. This particular laser design has a threshold current of about 8 mA for
the optical feedback due to the cleaved (uncoated) end facets (the reflection coefficient
is r :::::: 0.3). The light reflection can be improved through the lise of special reflecting
coatings on the end facets. In the latter case, the threshold current decreases below I mA,
as shown in Fig. 8.50 for the reflection coefficient r :::::: 0.8. The same figure illustrates
"'-,\
4J Lighl output ,
I
)
............... ~~ .............•
·····1···11····
••••• f t . •••• ~~
~ 1····1····11··
•••• •••• ••
,u~::U::~U:::~, 1111111111U11,
...."", ,,,,,,,,,,,,,,,,,, """" "",,,,,,,,"', ,,,,,,,,

,"""w ""..." ....." ....."""".....""",,"", ,-,
,,"", ,"" 1../11
""",,,,,,,, ...
,-,,-,,"""""""',"""",
,,,,,,,,....,,,,,,,,,,,,,,
,,,,,,,,,,,,,,,,"',,,,,,,,,,,,,,,,,,,,, ,,,,,,,,, "',,,,...,,,,,,,
,,...""',,,,""',,,,,,,,,,,.,....
......... ··.._ _ w'w~"""" ,,_'..',',.-.._
'''''''''',,'',,''',,',,'''''' """'",''''',,''''' ""'''"""",..
• metalliz<llionDactivc region
1-10 ~m
E~ll1ultilaycr !rnimplantcd region
Figure 8.51 Surface-emilling lasers.
also a typical output light power in the injection lasers: from 1 mW to tells ofmilliwans
depending on the particular diode and resonator design.
Surface-emitting lasers
Thus far, in considering quantum-wclliascrs we have assumcd that the laser is dcsigncd
for light propagation along the quantum.well layers. Another possible design uses
light propagating perpendicular to the layers in a so-called verricaJ geometry. The
amplification of light passing through a quantum-well layer can be defined as
Iou, - I in = p.
Iin
Here, I lIl and I OUl are the input and output light intensities, respectively. The quantity
fJ is estimated through the gain coefficient as fJ(w):::::: a(w)L, where L is the width
of the quantum well. It is easy to see that fJ is typically very small. For example, if
ao = 100cm- 1 and L = IDa A, we get fJ = 10-4 . In order to obtain laser oscillations in
a vertical-geometry structure, one should employ a multiple-quantum-well structure and
provide near-perfect mirrors with extremely high reflectivities. Figure 8.51 (a) depicts
schematically such a smj{lce-emirri"g lase,.. The laser design includes an active region
providing for high light gain, dielectric multi layers, metallic contacts. and implanted
regions. which form the light output. Layered dielectric mirrors give very high reflection
while the active region contains a multiple-quantum-well structure. The lateral sizes
of this laser can be reduced to the l-IO-~m range. A decrease in the surface area of
the diode leads to a considerable decrease in the magnitude of the threshold current.
In the case of the quantum-well structures just considered, we can assume a charac-
teristic current density of about 100 A cm- 2 ; accordingly, if the lateral sizes are each
5 J..UTI, the pumping surface area is 2.5 x 10- 7 cm 2 and the threshold current equals
25 flA.
Surface-emitting quantum-well lasers offer a new advantage of high-packing density.
Nowadays, technology allows one to fabricate an array of about 106 surface-emitting
Pb.Sn,_?- 1 - 1 - - - - - .
Pb n, ,Te t------......
PbS,?-, 1-1--
InJU..sb,_;r I I
Cd)ig,~Te 1-,--------11
.. I CCVb.
1~_...Ga...AsrPI-.r I I
Ga.1\s..5 bl_x I I
In.1\5..,P._.1 1-1-----<
(AI.Ca,_.l,In,_,As ~I-~
AI.Ca,_,As I-i
GaAII_"p.. t--t
In.Ca,_,As ~I------"
(AI.ca,_J,In,_1 I--i
CdS~Se l_z t---t
Cd..Zn._ ..S I I
, "1,,,,1 , "1,,,,1 ,I""
0.1 0.5 I 5 10 50 100
Wavelength (11m)
Figure 8.52 Speclrtll regions of lasers with various semiconductor Illedia. After P. L. Derry, A.
Yariv et lIl., "Ultralow-threshold graded·index separate-confinement single quantwn well buried
heterostructure (AI, Ga)As lasers with high rcfl«:livily coatings," Appl. Phys. Left.• 50,
1773-1775 (1987). Reused with permission from Pamela l. Derry. Amnan Yariv, Kum Y. Lau,
Nadav Bar-Chaim, Kevin Lee, and Jan Rosenberg, Applied Physics LeITers, 50, 1773 (1987).
Cll987 American Institute of Physics.
electrically pumped microlasers. Microlasers can operate at room temperature and

threshold currents are below 0.1 mAo
Blue and ultraviolet quantum-well lasers

As we emphasized previously, the wavelength and the energy quanla of light generated
by a semiconductor laser basically depend on the energy bandgap of the semiconductor
malerial selected. In Fig. 8.52, the spectral characteristics of the lasers from various
semiconduclors arc shown. For example, one can see that the most developed lasers
based on III-V compounds cover the wavelength range from approximately 0.6 ~m to
3 ~11l.
8.6 light~emitting diodes and lasers 301
Operation in the very interesting range of shorter wavelengths requires the use of
semiconductors with larger energy bandgaps. Well-known examples of such wide-
bandgap semiconductors arc representative II-VI compounds: CdS, CdSc, ZnCdSe,
etc. By lIsing different quanhlm-wcll designs, the blue injection lasers were realized
on the basis of these materials. However, the major problems associated with this class
of quantum-well lasers arc the device degradation and short lifetimes of the devices.
The electrical power dissipated in U~VI diodes is 100 high and causes rapid degra-
dation as a result of the generation of intrinsic defects. Much improvement must be
realized in this field in order for the group II-Vi-based devices to have wide practical
application.
Recently, another class of heterostructure materials has been studied in detail- the
nitrides of group Ill. It is difficult to obtain these materials in the usual single-crystal
form. However, thin nanometer-scale layers can be grown successfully with good quality
by applying the growth methods presented in Chapter 5. The nitrides of group ill include
InN, GaN, AIN, and all possible ternary and quaternary alloys. For InN and GaN, the
energy bandgaps are about 0.9 eV and 3.5 eV, respectively, while that for AIN is 6 eV.
Thus, by changing the In and AI content in the alloys In)'AlxGal_y_., ,one can increase
the bandgap from about I eV to 6 eV and realize lasers with wavelengths spanning the
range from 1.3 ~un (near-infrared light) to 0.2 Jl-m (deep-ultnlviolet light). The alloys
In.l'Gul_xN produce green, blue, and violet light.
light-emitting diodes
Although stimulated emission from the injection laser diode is very important, prnctically,
sub-threshold operation of the diode - when only spontaneous light is emitted - is in
many cases advantageous and has a number of applications. This mode of operation
does not require feedback to control the power output, which facilitates operation over
a wide range of temperatures, and is reliable and inexpensive. Diodes operating with
spontaneous light emission are called Iight-emitti"g diodes.
The important characteristic of the light-emitting diode is the spectral distribution
of emission. The spectrum of emission is determined, primarily, by the electronfhole
distributions over energy, which can be approximately described by corresponding Fermi
functions, as studied in Chapter 6. Thus, the ambient temperature, T, defines both the
spectral maximum ("the peak") and the spectral width of emission. The peak value of
the spectral distribution can be estimated as
The full width at half maximum of the distribution is I::!.w :::::: 2ku T /t/ and is independent
of w. In terms of the wavelength, A, we obtain
or
,
/>.). = 1.45)'~kB T. (8.46)
Yellow Ncar infrared

/,...------'-----~\
Ino 13 Gao. I ' ASo. l4 PO./lfi

""""
GaN"
/GaAs J h~30,2~ASD.1lO PO.40
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.\ 1.2 1.3
Wavelength "- (~J.m)
Figure 8.53 The spectra of Iighl-emitting semiconductor diodes with different bandgaps. Arter
Fig. 13 on page 696 of S. M. Szc, Physics o/Semiconductor Dm';ce.~ (New York, Wiley. 1981).
Reprinted with permission of John Wiley & Sons. Inc.
where Am corresponds to the maximum of the spectral distribution, 6>" and Am are
expressed in micrometers, and kB T is expressed in cY. Figure 8.53 shows the spectral
density as a function of the wavelength for light-emitting diodes based on various mate-
rials. The spectral density is normalized so that its maximum equals I for all samples.
For these different materials, the spectral linewidths increase in proportion to ),2, in
accordance with Eq. (8.46). From Fig. 8.53, one can sec that light-emitting diodes cover
a wide spectral region from the infrared - about 8 Jim for InGaAsP alloys - to the ncar
ultraviolet - 0.4 Jim for GaN. They are, indeed, very universalliglll sources.
Similarly to the case oflasers. the parameters of light-emitting diodes may be consid-
erably enhanced by using heterostructures, particularly quantum wells. Light-emitting
diodes may be also designed either in a swj{lce-emitting configuration, or in an edge-
emitting configuration. These configurations are illustrated in Figs. 8.54 and 8.55, respec-
tively. Surface-emitting diodes radiate from the face pamllel to the p-njunction plane.
The light emitted in lhe opposite direction is either absorbed by a substrate, or reflected
by metallic contacts. The edge-emitting diodes radiate from the edge of the junction
region. Usually, surface-emitting diodes are more efficient.
Since the diodes under consideration radiate through spontaneous emission, the spatial
patterns of the emitted light depend only on the geometries of the devices. Different
lenses can improve the emission pattern. Usually, the edge-emjtting diodes have narrower
emission patterns.
Light-emitting diodes find many applications, ranging from common lighting systems
to signal processing and light communications.
Unipolar intersubband quantum-cascade lasers

So far in this chapter, we have considered light emission and laser action based on
illIerband phototransitions involving both electrons and holes. Another type of photo
8.6 Lighl-emiUlng diodes and lasers 303
Rgure 8.54 A surface·cmitting diode.
Agure 8.55 An edgc-cmitting diode.
transition - illfrabolld absorption - is allowed in an ideal crystal system with hetcrojunc-

lions. Since the latter phototransitions arc drastically different from interband transitions,
laser action associated with these intraband transitions should differ in a fundamental
way from that studied for the laser schemes considered above. First of all, an intraband-
transition laser should employ only one type of carrier, i.e., it is a unipolar del/ice.
Second it should be based on electron transitions between confined states arising from
the quantization in semiconductor heterostructures. In order to create a population inver-
sion between two confined states, one needs to provide for (i) electron injection into a
higher lasing state and (ii) depletion of a lower lasing state. For this purpose, a vertical
scheme ofelectron transport has been proposed. This scheme is illustrated in Fig. 8.56(a).
The proposed heterostructure is a superlattice with a complex design for each period.
Each of the periods consists of four AlinAs barriers. forming three GaJnAs quantum
(a)
(b)
~ ... ::t:;;;:~_
...._.__S
FIgure 8.56 (a) Two periods of the 25-stage staircase coupled-well region of a quantum-cascade
laser under operational conditions. The laser phototransiliol1s arc indicated by arrows. They
occur between levels (subbands) 3 and 2 with the photon energy of295 meY. Lcvel 2
depopulates Ihrough levelland subsequent tunncling. The energy sepamtion between levels 2
and 1 is 30 mcY. (b) Energy dispersion for sllbbands 1, 2. and 3, pholotransitions, and
intcrsllbband·scattering processes (straight lines). Reprinted with pennission from J. Faist,
F. Capasso el oJ., "Quantum cascade laser," Science, 264. 553-556 (1994). Cll994 AAAS.
wells, and a graded AlLnGaAs region, which is doped. Under zero-bias conditions, the
overall band diagram resembles a sawtooth structure. Under an applied electric field,
the band diagram takes on a staircase structure as shown in Fig. 8.56(a). The barriers
form three coupled quantum wells with three quasi-bound levels. These three levels are
labeled in Fig. 8.56(a) by I, 2, and 3. Each of the confined slates originates from one
of lhe wells. The structure is chosen so that there is a considerable overlap between the
wave functions of the upper state, 3, and the intermediate state, 2. The same holds for
wave functions of states 2 and I. Under a voltage bias, the potential in the doped regions
is almost flat, as shown in Fig. 8.56(a). Thc electrons are injected from the doped regions
through the barrier in the confined state 3 of the first quantum well. From this state, they
relax primarily to state 2. There are two processes of relaxation: phonon emission and
photon emission. In Fig. 8.56(b). these processes are shown for electrons with various
values of tile in-plane wavevector, k. The three indicated subbands, EU.3(k), correspond
to the three confined states. The straight arrows represent intersubband phonon relax-
ation. The third confined state, 3, is selected to provide depletion of state 2 as fast as
possible. Thus, in this manner we have a three-level schcme whereby the upper level is
pumped by the direct injection of electrons from the doped region. The second level is
depleted due 10 strong coupling with thc lowest level I. From level I. electrons escape to
the next doped region. Then thc processes arc repeated in each subsequcnt period of the
superlartice. One can say that the carriers make transitions down through such a cascade
structure.
To describe the processes in the cascade structure we need to define the numbers of
electrons in three states: II], "2. and Ill. The criterion for population inversion between
levels 2 and 3 should be
We can write simple balance equations for,,] and 112:

dll] I 11]
- = --J-- (847)
dt e f]2
- ="3
dt
--- '" (848)
where J is the density of the injection current, and t]2 and f21 arc relaxation times
between the states 3 and 2, and 2 and I, respectively. In Eq. (8.48), we neglect the
inverse I ~ 2 process since state I can be regarded as almost empty as a result of
fast clectron escape to the doped region. For the steady-state conditions, we obtain the
conccntrations
I '21
11] = --Jt]2 and 112 = "]-.
e
and population invcrsion
'"
fj.n=II]-1I2=-~Jf]2(1- f 21 ). (849)
e tn
Thus, to create a population inversion, one should design the laser so that
(8.50)
In order to fabricate such unipolar laser structures with vertical electron transport, very
precise sophisticated semiconductor technology is necessary.
For a particular device structure (Fig. 8.57) with an optical path of about 700 J.1m and
the mirror reflectivity"l ="2 = 0.27, the laser output-current characteristics at various
temperatures are shown in Fig. 8.58. The insets of Fig. 8.58 show the current-voltage
characteristics and temperature dependence of the laser threshold current. The laser
threshold currcnt can be approximated by Ilh = C cxp(T/112), where the constant C is
about 900 rnA and T is measured in degrees Kelvin. From Fig. 8.58, it follows that the
output power reaches tens of rnilliwatts.
The emission energy is in the range 275-310 meY. Spectra of the laser output for
various currents at T = 80 K are presented in the right-hand inset of Fig. 8.58. For
this case, the threshold current is about 1.06 A. This inset clearly demonstrates a sharp
narrowing of the emission spectra above the laser threshold: the spectra reduce to a sharp
peak at I = 1.1 A > lib,
Thus, the unipolar cascade laser is drastically different from the lasers based on
intcrsubband phototransitions. The properties of intraband phototransitions and, con-
sequently, the properties of the unipolar laser are determined to a large degree by
quantum confinement; accordingly, this novel laser can be tailored for operation in
GalnAs (Sn-doped) 11= 2.0 x l(yocm-J 20.0 om

-
< .>-
GaillAs 1.0 x 10 18 670.0 "'"
- >.
C'"
0-
AIGalnAs (graded) 1.0 x 10 18 30.0 o
AlinAs 5.0 x 10 17 1500.0
AllnAs 1.5 x 10 17 1000.0

AIGainAs (digitally graded) I.SxIOI7 18.6
Active region undoped 21.1
GalnAs 1.0 x 10 17 300.0
8"
~
AIGalnAs (digitally graded) 1.5 x 10 11 14.6

AIGalnAs (digitally graded) 1.5 X 10 17 18.6 T~"
Active region undopcd 21.1 x f5 ~
GalnAs 1.0 X 1017 300.0

+ ~
AIGalnAs (digitally graded) I.SxIOI7 33.2
AllnAs 5.0 x 10 17 500.0
Doped o+-lnPsubstratc
Figure 8.57 A schematic cross*scction of the cascade laser structure. The whole structure
consists of 500 layers. Reprinted with pcnnission from F. Capasso, "Quantum cascade laser,"
Science, 264, 553-556 (1994). It' 1994 AAAS.
the spectral region from the middle infrared to submillimeter waves. Quantum cas-
cade lasers are very sensitive to the ambient temperature and typicalJy work at reduced
temperatures.
In conclusion, we have shown that nanostructures playa key role not only for down-
scaled electrical devices, but also for optical devices. They facilitate improvements in
bipolar injection lasers and make it possible to realize cascade-laser structures, which
work on the basis of innovative concepts.
8.7 Nanoelectromechanical system devices
So far, we have concentrated on the electron properties of nanostructures and have

shown that electronic effects on lhe nanoseale can be exploited for electrical devices.
Mechanical properties ofnanostructurcs are very different from those of bulk samples.
One can use both electronic and mechanical properties on the nanoscale to develop a new
class ofdevices - nanoelectromcchanical devices. Fabrication of nano·electromechanical
8.7 Nanoelectromechanical system devices 307
"
::i ,.
• ./ 10 K • _11 ' ..
we .........
<La • ..,. '"
I."
->:
I
o,so
50 rMoo !
-
E
BOd
0- !
a·
.. i
5
SiaM '" '" - ........'" '"
0.250.5 1.0 1.5 2.0 2.5 3.0

Current (A)
FIgure 8.58 The measured optical po\\i'eT P from a single facet of the quantum-cascade laser with
the structure presented in Fig. 8.57 and an optical cavity length of 1.2 rnm. The resuhs are given
for four different temperatures. The insets on the letl show the dependence of the laser threshold
curren! as a function oftcmperature and the current-vollagc characteristics of the device. The
inset on the right orthe figure shows the output spectrum for a 1.2-mm-long laser below and
above threshold (/111 = 1.06 A) al 80 K heat-sink temperature. Above 0.4 A the luminescence
peak was found 10 grow 5uperlinearly with current due 10 optical gain causing the line to narrow.
After F. Capasso. "Quantum cascade lasers: a unipolar intersubband semiconductor laser," in
Proceedillgs o/the Imemarionul Conference ol/the Physics o/Semicollductors (Singapore,
World Scientific, 1995). pp. 1636-1640.
systems (NEMSs) was described in Section 5.9. Here, we will discuss a few particular
NEM devices.
Resonators. Parametric amplification

Parametric resonator NEMSs serve as mechanical amplifiers in a narrow frequency range.
The basic idea ofa mechanical parametric resonator can be illustrated by the following
simple example. Consider a simple parallel-plate capacitor, in which one plate is a part
of the mechanical resonator. while the second plate is fixed, as shown in Fig. 8.59. A
=,
displacement, of the resonator plate changes the spacing in the capacitor and thus the
device capacitance, C = EOSj(d + =), where S and d are the area of the plates and the
equilibrium distance between them, respectively. Since the displacements are small, we
can use the approximation
C '"
"0
<
d
S( d-+ d'-, ...).
I-
--
If a time-dependent voltage V(I) is applied to the capacitor, its electrical energy becomes
I ,
E"=2 CV (t).
Allure 8.59 A mechanical parametric resonator fabricated by the SIMOX technique.

Characll~ristic dimensions of the resonator are indicated. From A. N. Cleland, Foundations of
Nallomecholl;cs (Berlin, Springer-Verlag. 2003).
Let M and K be the mass and the spring constant of our resonator. respectively. Then,
Newton's second law for the resonator plate takes the form
d2= dz d£rl
M-+My-+K:=f=- (8.51)
dl' dl d:
Here, we introduce the term My dz/dl describing the damping of the resonator. The
force acting on the resonator is
f = - 'oS
- V , (I) ( 1+- ;
(fl d
2:) (8.52)
i.e., it is proportional to the square of the driving voltage, V(t). Importantly, the force
contains a term proportional to the displacement z. We can characterize the resonator by
the frequency of vibration, Wo = J KIM, and the quality factor. Q = woly.lntroducing
the notation
2.05 V'(t)
!>K(I) = dl (8.53)
and
.05 V'(I)
Fo(1) = - d' . (8.54)
we can rewrite the force equation as
d':
-,
dt-
+ WO d:
- -
Q dl
+ (
wi + !>K(t»)
M
:= Fo(t). (8.55)
The physical significance of each of the quantities 6 K(t) and Fo(t) is obvious: 6KCt)
describes a parametric modulation of the spring constant by an applied voltage. while
FoCt) is the displacement-independent driving force. The latter equation is the simplest
basic equation for the effect ofpararnctric resonance: the parameters 6K and Fo have the
same time dependence; i.e., they are both proportional to V2(t).lfthe voltage depends on
time harmonically, V ex: sin(wt), we find that both the driving force and the spring con-
stant oscillate with the driving frequency, which varies with time as cos(2wt). In general,
a NEMS can be fabricated with two or more capacitor plates; then, different voltages
can be applied to these capacitors. Thus, to generalize the analysis we may assume
that the spring-constant modulation, 6K, and the driving force, Fo , have independent
modulation frequencies, Wp and Wo:
6K = 6Ko sin(wl't);
FD = FOCDS(WDI +¢).
Here, <p denotes a phase shift between these two dependences. More specifically, the
basic equation now becomes:
2
d z (2
---:; + -WO dz
- + Wo + 6Ko sin(wpt) ) z = Fo COS(Wot + <;b). (8.56)
dl- Q dl M
This equation is the so-called Mathieu equation with damping. Analysis of this
equation yields the following results. First, let the parametric modulation be absent
(6K o = 0), then the equation describes vibrations of a resonator with frequency cvo.
For small damping (Q »
I), vibrations will manifest a strong resonance at the driving
frequency, Wo ~ wo:
Z(I) = A CDS("'OI) + 8 sin("'01). (8.57)
FoQ . FoQ
A = K smt/>. B = K cost/>. (8.58)
This result shows that, for an oscil1ator with a quality factor Q -l> 00, the vibration
magnitude diverges.
If a parametric modulation is tuned by a small driving force, additional strong reso-
nances arise at WI' = 2wo/ fI, with II being an integer. That is, the parametric modulation
resonates for all submultiples of the frequency 2wo. Consider, for example, the case of
/I = I, i.e., WI' = 2Wo and WD = WOo In the limit of high Q, it is possible to find a solution
in the form of Eq. (8.58) with

FoQ sin ¢ FoQ cos ¢
A - '-':;=- :---;.;=,-;;-=;;,. 8 - '-"f"- --,-;;;=':':~=;: (8.59)
P - K 1 - Q D.Ko/(2K) P - K 1+ Q D.Ko/(2K)
Thus, the oscillations depend essentially on the parametric coupling parameter 6. K o,
and the amplitudes of the mechanical vibrations can be controlled by these parameters.
By comparing these results with and without parametric coupling, i.e., Eqs. (8.59) and
(8.58), we can define the parametric gain,
G- j A'P + 8'P
- ,fA' + 8"
Using the above formulas for the coefficients A. B. A p , and Bp, we find
1/
cos 2 = nl2
-----
0 ---- .. _
0 . 1 . . , . .(0=
. _
,
0.01 0.1 1
Q l1KoI(2K)
Figure 8.60 The gain/loss coefficient G as a function of the quality factor Q for different phase
shifts between spring-eonstant modulation and the driving force. From A. N. Cleland
FO/mdaliolls o/Nanomechallics (Berlin, Springer·Verlag. 2003).
For I:!J. K '# 0, the gain depends on the relative phase ¢. If 4> = O. rr. 2rr..... the ampli-
tudes of the vibratjons arc suppressed and G < I. If ¢ = rr/2. 3rr/2. 5rr/2 ..... the
vibrations are amplified and G > I. Formally, if I:!J.K o -+ 2K / Q, the amplitude diverges
with G -+ 00. This phenomenon is calledpllramerric resonance. Both de-amplification
and amplification regimes are illustrated by Fig. 8.60. We can conclude that, if the quality
factor, Q, is large, an amplification of the amplitude of vibrations can be obtained even
for smallmoduJations of the driving force.
The resonator of a micromechanical parametric oscillator, shown in Fig. 8.59, is
fabricated by the SIMOX technique. The dimensions of the suspended resonator plate
are 4 ~m x 4 ~un x 0.2 ~m. The substrate is grounded. A metal is deposited onto
the resonator plate. The resonator is driven by a voltage applied between the platc
and the substratc. The fundamental frequency of the resonator is wo/(2rr) = 485 kHz.
l11Us, the amplified signaJ is at this frequency, while the parametric drive is at WI' = 2W(l.
111ese parameters correspond to the lowest resonance in the Mathjeu equation (8.56). To
detect and measure displacements, one can use reflection of a laser beam from the
substrate and lhe resonator. Interference of these two reflected signals yields good dis-
placement sensitivity. In Fig. 8.61, the measured square of the oscillation amplitude is
presented as a function of the pump amplitude for such a parametric oscillator. The
maximum amplification, G, achieved is about 10.
l1le resonance properties ofNEMSs will undoubtedJy be employed in a broad range of
applications. Obviously, one of the principal areas will be signal processing in the very·
high-frequency (VHF), ultra.high.frcqucncy (UHF), and microwavc·frequcncy bands.
Mechanically detected magnetic resonance imaging

Another promising application ofNEMSs is mechanically detected magnetic resonancc
imaging (MRJ).1t is well known that the phenomenon of nuclear magnetic resonance is
widely used for diagnostic purposes in medicine. The conventional inductive detection
10r---~--~--~--
,../
1 ~....
_i - - ~--
~ - - • response
---fit
o.0.00
1 ':-::-~---:--=----:--':":"-----:-"-::-----:-
0.05 0.10 0.15 0.20
Pump amplitude (V)
Rgure &61 The square of the oscillation amplitude ("the power of oscillations") versus the
pumping amplitude for the parametric resonance oscillator shown in Fig. 8.59. From A. N.
Cleland Foundations ofNanomechanics (Berlin. Springer-Verlag. 2(03).
displacement sensor
(fiber optic interferometer)
force delector
t:::::~:~:IN;E;MSresonator)
gradienl magnent
(nanoscale magnet
excll.llon llejd
surfaces
of constant
__- ~;:\3:~~-::;__')I~(~rl~,:on or mlcrostrlp)
magnetic field
resonant ...._-1.,
"slice"
Rgure 8.62 The principal scheme of mechanically detected magnetic resonance imaging. After
M. L. Roukes. "Nanoelectromechanical systems:' in Technical Digest ofthe ]000 Solid-State
and Actllator Workshop, Hilton Head Island SC. pp. 1-10.
techniques still take about 10 14_10 16 nuclei to generate a measurable signal of the mag-
netic resonance. This means that state-of-the-art MRI in research laboratories attains,
at best, a maximal resolution (minimum voxel size) of the order of I J.UTt. For standard
clinical MRI, the resolution yielded by commercial instrumentation is much poorer, with
a voxel size of aboUl I mm.
Mechanically detected MRJ, now commonly called magnetiC resonance force
microscopy (MRFM), is significantly more sensitive than conventional MRI. There are
several principal components ofa MRFM instrument, which are presented in Fig. 8.61.
An antenna structure in the form of a coil or a microstripline generates a radio-frequency
field of frequency woo The static magnetic field created by a miniature magnet splits the
spin states of impurities within a sample and provides a resonance interaction of the
VSD Iso
Figure 8.63 The electron shuttlc - a dcvice with electron transport providcd through mechanical
motion oflhe cantilever. C is a nanosizcd cantilever. Sand D are the source and dmin contacts.
and G I and G2 are the gate contacts, which move the cantilever. The inset depicts the electric
circuit of the structure. From A. N. Cleland FOl/nd(f/ions ofNanomechanics (Berlin.
Springer·Verlag, 2003).
radio-frequency field with these spins. The magnet is attached to a mechanical resonator
of a cantilever type. The interaction of the resonant spins with the magnet results in
a time-varying back-action force upon this cantilever. This force can be detected by a
sensor with high resolution of displacements, such as an optical interferometer. All of
these clements constitute a resonant force sensor. Such a sensor can dctcct the extremely
weak forces exerted by the resonant spins upon the mechanical system. Microscopy is
realized by scanning the cantilever with attached magnet over the sample. By correlating
the resonant mechanical response with the cantilever position, one can obtain spatial
imaging of spin density.
The resonance properties ofNEMSs provide key advances for MRFM. Indeed, utiliza-
tion of a nanometer-scale ultra-high-frequency mechanical resonator makes it feasible to
couple directly the spin precession and mechanical vibrations and to improve drastically
the resolution ofMRJ.
The field of MRfM is still very much in its infancy. Sustained efforts arc required in
order to take it from a scientific demonstration to a useful technique for high·resolution
MRI. With its potential for atomic resolution, such efforts arc of great potential impor-
tance, especially for biochemical applications.
Coupling of electron transport and mechanical motion. The electron shuttle

The coupling of electron transport and mechanical motion of NEMSs gives rise 10 new
effects, which can be useful for a number of applications. Consider an example of such
a coupling, which can be called the electron sll/mle. The structure contains a metallized
cantilever suspended between two metallic electrodes, as shown in Fig. 8.63, where an
45
; 300 K
~
-<c
-
~
-;
0 50 100
((MHz)
~ 30
~
...
~
~4
15 - -
-a
2
1.0 2.0
v (V)
o 50 100
J(MHz)
Figure 8.64 Electric current in the electron shuttle versus frequency in the gate circuit. Upper
inset the cantilever is blocked and there is no currenl. Lower inset: the shuttle current as a
function of the source-drain voltage at a given frequency in the gate circuit. From A.
Cleland Fou"dations ofNlIllomechallics (Berlin. Springer- Verlag, 2003).
elcctron micrograph of the electron shuttle is presented. Two gates are fabricatcd to
drive thc cantilcver electrostatically. Thc cantilevcr can be driven by the voltage, VG1Gl'
so that the cantilever can nearly contact each of the spatially separated electrodes. A
voltage VSD is applied to the electrodes, which can be considered as a source and a
drain. If the frequency of the voltage applied to the gates, VG,Gl' does not coincide
with the resonance frequency of the cantilever, the amplitude of cantilever vibrations is
small. Accordingly, the cantilever does not mechanically contact the source and drain
electrodes, and electron transport is suppressed. As soon as the frequency VGIG ! matches
the mechanical resonance, the cantilever contacts the source and drain during each cycle
of the mechanical motion. When the cantilever contacts the electrodes successively, the
metallized end of the cantilever charges and discharges, and transfers electrons betwecn
the source and the drain. In Fig. 8.64, results of measurements ofthe source-drain current
are presented; several peaks in the current correspond to different resonant vibration
modes of the cantilever. Only for these resonances are the amplinldes of the vibrations
large enough to support electron transfer through tbe nanostructure. If the motion of the
cantilever is blocked, no current is observed, as seen from the upper inset to Fig. 8.64.
The lower inset shows the magnitude of one of the current peaks as a function of VSD
at a given resonance frequency. Thus, this NEMS actually presents an electron shuttle.
The number of electrons transferred depends on the dimensions of the metallized island
Figure 8.65 A sel of SiC NEMSs. Submicrometcr double·c1ampcd SiC beams exhibit
fundamental resonance frequencies from 210134 MHz. From M. L. Roukes,
"Nanoeleclromechanical systems," in TecJmicol Digest o/the 2000 Solid·Srate alld Actuator
Workshop, Hilton Head Island., SC, pp. 1-10.
on the cantilever. By decreasing the size of tlle island, it becomes feasible to transfer a
single electron at a time.
Following these examples ofNEMSs with different functions, we consider the basic
parameters that determine the performance of any NEMS. Actually, these parameters
are practically the same as for an electronic device: the response time or characteristic
frequency, the quality factor (energy loss), the operating power, the signal-to-noise ratio
(sensitivity), etc.
Frequency
We start by considering a NEMS as a mechanical vibrator at a natural angular frequency,
WOo The frequency can be approximated by wo = J K/ M, where K is an effective spring
constant and M is the mass of the mechanical vibrator. Underlying these simplified
"effective" terms is a complex set of elasticity equations that govern the mechanical
response of these objects. If we reduce the size of the mechanical device while preserving
its overall shape, then the fundamental frequency, WOo increases as the vibrator's linear
dimension decreases. This is obvious because the mass is proportional to the volume
of the mechanical vibrator, while the effective spring constant for flexural deformations
increases with the decrease of its length. This increasing frequency effcct is important
because a high frequency response translates directly to a fast temporal response to
applied forces. It also implies that a fast response can be achieved without the expense of
making stiffstructures. Moreover, a small spring constant provides very high mechanical
responsivity.
The shape of the vibrations and resulting frequencies depend on the way the beams
arc clamped. Three variants can be realized in a particular device: (i) both ends clamped
or free. (ii) both ends pinned, and (iii) a suspended beam clamped at one end (the can-
tilever). In Fig. 8.65, images of double-clamped SiC beams are shown. The beams are of
Table 8.2 Estimated fundamental frequency versus geometry for SiC, Si, and GaAs mechanical resonators
Resonator dimensions
L~ x L y x L: in (,"un)
Boundary conditions 10xO.2xO.1 I x 0.05 x 0.05 0.1 x 0.01 x 0.01
Material SiC, Si, GaAs SiC, Si, GaAs SiC, Si. GaAs
Both ends clamped or free 12,7.7,4.2 MHz 590,380,205 MHz 12,7.7,4.2, GHz
Both ends pinned 5.3. 3.4, 1.8 MHz 260, 170,92 MHz 5.3.3.4. 1.8 GHz
Cantilever 1.9, 1.2,0.65 MHz 93,60, 32 MHz 1.9, 1.2,0.65 GHz
After M. L. Roukcs "Nanoclectromechanical systems", in Technical Digest ollhe 1000 Solid-Slate Sensor alltl
Actulltor Workshop, Hilton Head Island, SC, pp. I~IO.
different dimensions and exhibit fundamental frequencies of flexural vibrations varying

from 2 MHz to 134 MHz. Table 8.2 displays frequencics for the fundamental jlexu-
ral modes of thin beams for various materials (SiC, Si, and GaAs) and dimensions
spanning the domain from micromechanical systems to nanosystems well within the
nanoregime. The last column in Table 8.2 represents the dimensions currently attain-
able with advanced electron-beam lithognlphy. The materials SiC, Si, and GaAs are of
particular interest, because they are available with extremely high purity as monocrys-
talline layers in epitaxially grown helerostructures. The numbers in Table 8.2 are rough
averages for the various commonly used crystallographic orienlations. It is particularly
notable thai, for structures of the same dimensions, Si yields frequencies of a factor of
two, and SiC of a faclor of three, higher than those obtained with GaAs devices. This
increase reflects the increased phase vclocityofsound as well as the higher stiffness of the
materials.
Thus, mechanical resonators with fundamental frequencies above 10 GHz (10 10 Hz)
can now be built using surface nanomachining processes involving state-of-the-art nano-
lithography on the 10-nm scale.
Quality factor
Another important parameter, which characterizes the rate of energy dissipation in a
NEMS, is the so-called quality faCIOI', Q. If we define a decay time of the flexural
vibrations of the beam in a NEMS, rd, then, Q::::::: WOrd. The quality factor achieved for
a NEMS in a moderate vacuum is in the range from I(}3 to lOs. This greatly exceeds the
quality factor typically realized for electrical (microwave) resonators. The small degree
of internal energy dissipation in a NEMS directly relates to the high attainable force
sensitivity and low operating power levels.
Characteristic operating power level

Applications of NEMS resonators involve the use of a specific vibrational mode, typi-
cally one of the lowestjlexuralmodes. A rough understanding of the minimum power
Table 8.3 Representative operating power levels for NEMSs

(1 .now.n = 1 .W = 10- 18 W)
Frequency, Wo Quality factor. Q Minimum power. Pmin
100 MHz 10000 40aW

100 MHz 100000 4aW
I GHz 10000 400aW
I GHz 100000 40aW
After M. L. Roukes (2000) "'NanocleetTomeehanical systems," in

Technical Digest ojlhe 2000 Solid-Stare Sensor and Actllator Work-
shop, Hilton Head Island, SC, pp. 1-10.
necessary to operate with a NEMS using the ftexuralmode can be obtained as follows.
At equilibrium the average energy of such a mode is equal to the thermal energy kB T.
To perform an operation, the energy of the external input signal should be larger than the
thermal energy. The characteristic time scale for energy exchange between the mode, at
frequcncy W(), and its surroundings is just the decay time, 'd = Q/wo. Thus, thc mini-
mum powcrofthe signal which has to be applied to the system to drive it to an amplitude
above the thermal fluctuations is
P> Pmin = kBT.
In Table 8.3. we present values of Pmm for various frequencies and quality factors at room
temperature. As displayed in Table 8.3. this minimum power can be extremely small for
a NEMS. For device dimensions accessible today via electron-beam lithography, the
characteristic level is of the order of tens of attowans (10- 17 W!). This is many orders
of magnitl.lde smaller than Ihe power dissipation in contemporary systems of similar
complexity based on digital devices Ihal work solely with clectric signals.
Dynamic range of a NEMS

From Table 8.3 it is clear that NEMSs have the potential to provide new types of ultra-
low-power electromechanical signal processing. However, realization of these potential
advantages is not a simple task. To utilize the full potential of a NEMS, displacement
transduction schemes should be capable of providing resolution of the beam displace-
ment at the level of Ihe thermomechanical fluctuations. Indeed, at a finite temperature
therc aJways occur mechanical fluctuations and, thus, there exist small random (chaotic)
displacements. 6IT, of the bcam. These displacements can be readily estimated by
equating the potcntial cnergy of our vibrator, ~K(6zr)2, to the thermal energy kBT.
which gives the amplitude of such thermally induced vibrations as 6.=r = J2k B T I K.
As we indicated above. the spring constant K decreases as the dimensions of the beam
diminish. Thus, thermally induced displacements scale with length, L as 6zr ex 1/ JL.
I
8.8 Quantum-dot cellular automata 317
Simultaneously, the signal amplitude of vibrations for a mechanical device (in a linear
operation regime) scales downward in direct proportion to its size. It is important to esti-
mate the acceptable level of operatjonal displacements of the sllspended slructure, i.e.,
the so-called dynamic range for a linear NEMS. To estimate the characteristic dynamic
range, we need to define the displacement amplitude corresponding to the onsel of non-
linearity. This nonlinearity implies that, in a power-series expansion for elastic potential
energy, the largest term beyond the quadratic term (i.e., beyond the Hookes'-Iaw term)
become-s important. Fora double-clamped beam, this condition translates into the relation
.!lZN ..... 0.5L=. This criterion depends only upon the beam thickness, L=, in the direction
of vibmtions. The linear dynamic range for a NEMS can be defined as the ratio of the
nonlinearity onset, .!lzN, to the thermal displacement• .!lZT: DR 0:; 10 In(.6.=N/.!lzr).
The following example highlights the characteristic displacements and the dynamic
range in a relatively small NEMS. Consider a suspended Si structure of dimensions
0.1 ~1 x O.Olllm x 0.01 J.un with a quality factor Q = 104 at T = 300 K. Then, we
find .!l=r :::::: 0.09 nm and .6.z N :::::: 5 11m. For these parameters. a high value ofthe dynamic
range, DR, of approximately 40 is obtained.
These considerations and estimates indicate two verycruciaJ areas for EMS engineer-
ing necessary to provide femtowatt to picowatt regimes: (i) development of ultra-sensitive
transducers that arecapableofenhanced displacement resolution with increasingly higher
frequencies as device sizes are progressively scaled downward and (ii) development of
techniques tailored to operate over the entire dynamic range of the NEMS. As we studied
previously, nowadays, probe microscopy techniques are capable of probing and measur-
ing quantitatively estimated ultimate displaccments.
In conclusion, a nanoclcctromechanical system consists of a nanomcter-to·
submicrometer-scale mechanical resonator that is coupled to an electronic device of
comparable dimensions. The mechanical resonator may have a simple geometry, such
as a cantilever (a suspended beam clamped at one end) or a bridge (a suspended beam
clamped at both ends) and is fabricated from materials such as silicon using litho-
graphic and other techniques similar to those employed for fabricating integrated cir-
cuits. Because of their submicrometer and nanoscale size, the mechanical resonators can
vibrate at frequencies ranging from a few megahertz up to around a gigahertz. The qual-
ity factors of these resonators grcatly exceed those of typical microwavc resonators. A
NEMS operates with low power dissipation in a wide dynamic range. These properties of
NEMSs open the way to a number ofapplications, ranging from signal processing to novel
detectors.
8.8 Quantum-dot cellular automata
As we discussed previously, the general tendency is that improvement in technology

leads 10 progressive scaling down of electronic devices and widening of their func-
tionality. However, as more and more devices are placed into the same area, the heat
generated during a switching cycle can no longer be removed and this may result in
31 a Nanostructure devices
o- cell - - physical intcraction
Agure 8.66 A schematic picture of a cellular array with intcrconncctions via physical interaction.
considerable limitation of the speed of operation. In addition, interconnections betwcen

devices can not be scaled in accordance with lhe device scaling because of the effects
of wire resistance and capacitance. The latter phenomenon can lead to a wiring bot-
tleneck. In such a situation, alternative approaches arc needed in order to solve these
problems.
Contemporary nanotechnology can support alternative electronic devices and system
architectures. One such approach is based on quantum dots arranged in locally inter-
connected cellular-automata-like arrays. The fundamental idea of quantum-dot cellular-
automata operation is to encode information using the charge configuration of a set of
quantum dots. Importantly, in the quantum-dot cellular-automata approach, the infor-
mation is contained in the arrangement of charges of the dots, rather than in the flow
of the charges (i.e., current). It can be said that the devices interact by direct Coulomb
coupling rather than via the current through the wires. Figure 8.66 illustrates a locally
interconnected array of cells. Obviously, a dense arrangement of nanometer-scale quan-
tum dots can provide the necessary physical interactions inside the array. These physical
interactions between the elements, together with the topology of the system. determine
the overall functionality of the array.
As discussed in Section 5.5, technologies for the growth and processing of quantum-
dot arrays are already available. They facilitate the fabrication of quantum dots with the
necessary properties, arrangement, etc.
In quantum-dot cellular automata, the building block is called a cell. Figure 8.67(a)
shows such a cell. A cell is composed of(at least) four quantum dot's positioned at the
corners of a square. A cell contains hvo excess electrons, which arc allowed to tunnel
behveen neighboring quantum dots in the cell. Tunneling out of a cell is assumed to be
completely suppressed by the potential barriers behveen the cells.
8.8 Quantum-dot cellular automata 319
(a) (b)
(I) 0 <*l 0 0 ~
0 0
0 0 <*l <*l C
0 (I) P=-I P~+I
Rgure 8.67 A quantum-dot cell consisting of five dots with t\\'Q clei:trons in the cell: (a) the
Coulomb repulsion cau~s the electron 10 occupy the antipodal sites: and (b) ruiO bistable states
result in different polarizations. From S. Craig. P.. lent el aJ.• "Bistable saturation in coupled
quantum dots for quantum cellular automata:' AppJ. Phys. Left., 62. 714 (1993). Reused with
permission from S. Craig, P. Lent, D. Tougaw, and W. Porod. Applied Physics Letters, 62. 714
(1993). 1993 American Institute of Physics.
1.0 r-~~""""'~~f7",----"""
0.5
•
• ~
0
c(" 0.0 0
0
cell 2 cell I
-0.5
I
-1.0 L-_~_ _L'..L....I~~..........~..........J
-1.0 -0.5 0.0 0.5 1.0
Pz
Agure 8.68 The cell--eell response. A polarized cell I implies the polarization of cell 2 due to the
Coulomb repulsion of the electrons populating the cells. The strong nonlinearity of the cell-eell
response plays the same rolc as the gain in a conventional digital device. From S. Craig. P. lent
et al.. "Bistable saturation in coupled quantum dais for quantum cellular automata." AppJ. Phys.
Lett., 62. 714 (1993). Reused with permission from S. Craig, P. lent, D. Tougaw. and W. Porod,
Applied Physics Letter.f. 62. 714 (1993). C> 1993 American Institute of Physics.
The Coulomb repulsion between the electrons in a cell tends to place them at antipodal
sites in the square. For an isolated cell, there are 1\\'0 energetically equivalent arrange-
ments of the extra electrons, which are denoted as "cell polarizations:' P = + I and
P = - L The polarization is used to encode binary information. Forexample, if P = + I
represents a binary I, then P = - I can represent a binary 0. The two polarization states
of a cell will not be energetically equivalcm if another cell is nearby. Figure 8.68 shows
how one cell is influenced by the state of the neighboring cell, The inset illustrates two
cells where the polarization PI is determined by the polarization of the neighbor, Pl •
(a) (c)
,.........fixed driver
1~~~~~~~1
wire corner
I
(b)~
~
l~~~~
fan-ow ~
~
I
Figure 8.69 Examples of the simplest quanlum-dot cellular-automata armys: wire, corner,
fan-out, and inverter. From Y. Ono, A. Fujiwara el al., "Manipulation and detection of single
electrons for future information processing," J. Appl. Phys., 97,031101 (2005). Reused with
permission from Yukinori Ono, Akira Fujiwam, Katsuhiko Nishiguchi, Hiroshi Inokawa. and
Yasuo Takahashi, JOlll'llul ofApplied Physics, 97, 03 I 101 (2005). (I;') 2005 American Institute of
Physics.
through the interaction of the electrons in the cells. Let us assume that the polarization
Pl has a given value corresponding to a certain arrangement of the charge in the cell 2.
Owing to the charge repulsion, the response of cell I is given by the strongly nonlinear
dependence presented in Fig. 8.68: a small asymmetry of charge in cell 2 is sufficient
to break the degeneracy between the two possible states of cell I and leads to the same
configuration of cell I. Thus, in some cell arrangements, fixing the polariZ:1tion of a cell
at an edge by an external bias determines the polarization of the cells at other edges. The
polarization is dependent on the cell configuration.
In the quantum-dot cellular-automata approach, the circuit is built by forming a
tree of cells. Figure 8.69 shows some elements of the quantum-dot cellular automata.
Figure 8.69(a) is a binary wire. The polarization of the leftmost cell is fixed, which rep-
resents the input. Then, all the other cells, including the output cell, align with the same
polarity because it is the most energetically favorable. Flipping the polarity of the
input cell results in the flipping of all the other cells. During this procedure, no
direct current flows in the circuit. Cells positioned diagonally from each other tend
to anti-align. This feature is employed to construct other logical clements. A fan-
out, a corner, and an inverter arc shown in Figs. 8.69(b), 8.69(c), and 8.69(d),
respectively.
In conclusion, quantum-dot cellular-automata systems exploit the interactions between
quantum dots on the nanoscale. TI1CY arc able to perform all processes necessary for signal
processing. No current is used in the circuits built on the basis of quantum automata. The
quantum-dot cellular-automata approach presents a nanoscale alternative to conventional
microelectronics and nanoelectronics.
8.9 Closing remarks
In this chapter we have considered several different nanostructure devices. It is convenient

to classify these devices into two categories. The first are the devices which are based
on usual classical principles of operation but considerably scaled down. This scaling
down facilitates the improvement of device performance and solves some fundamental
difficulties, which can not be overcome in microelectronics. These devices include field-
efTectlransistors and bipolar transistors.
Another category is nanostnlcture devices based on new physical principles. Among
these arc resonant-tunneling diodes - the simplest quanhllll devices. Since these devices
have nanoscale dimensional features, they have extremely short transit times for carrier
transport through the structures, which results in the possibility of generating ultra-high-
frequency electromagnetic oscillations up to the terahertz frequency rnnge. As exmnples
of new principles for three-terminal electronic devices (transistors), we analyzed the
quantum-interference and hot-electron trnnsistors, which have great potential for ultra-
high-speed operation.
To illustrate the contribution of nanostructures to optical devices, we have focussed
on the examples of quantum-well and quantum-wire lasers with bipolar injection, and
the multilayered quanhlm-cascade laser. The latter is a monopolar device, i.e., it uses
only one type of carrier (electrons). Exploitation ofnanostnlctures results in a dramatic
decrease of the threshold current necessary for laser generatjon and in a widening of
emission spectra.
We have analyzed some nanoelectromechanical systems with mechanical resonators
that can vibrate at frequencies ranging from a few megahertz up to 1 GHz and have
quality factors greatly exceeding those of typical microwave resonators. This implies that
nanoelectromechanical systems operate with low power dissipation. These properties
open the way for a number of applications, ranging from signal processing to novel
sensors/detectors.
Finally, we have described innovative systems for signal processing based on quantum-
dot cellular automata. These automata exploit interactions between quantum dots on the
nanoscale. This approach presents a nanoscale alternative 10 conventional microelec-
tronics and nanoelectronics.
The description of nanostructured devices which we presented here is not com-
plete. More information on traditional scaling down to nanosize devices, double-barrier
resonant-tunneling diodes, and lasers can be found in the following books:
M. Shur, Physics of Semiconductor Devices (Englewood Cliffs, NJ, Prentice-Hall.
1990).
S. M. Sze, Nigh-Speed Semiconduc1or Devices (New York, John Wiley & Sons, Inc.,
1990).
V V Mitin, V A. Kochclap, and M. A. Stroscio, QUllIl1//111 Heteros1rllc1/1res (New
Single-electron transport is described in the following references:

K. K. Likharev, "Correlated discrete transfer of single electrons in ultrasmall tunnel

jllllclions;' IBM 1. Res. Develop., 12, 144 (1988).
M. A. Kastner. "The single electron transistor," Rev. Mod. Phys., 64, 849 (1992).
Y. Ono, A. Fujiwara et (1/., "Manipulation and dctection of single electrons for future
information processing," 1 App/. Phys.• 97,031101 (2005).
A very detailed review of nanoelectromechanical systems is given in the text
A. N. Cleland, Foundations ojN(lIIomecJumics (Berlin, Springer-Verlag, 2003).

Appendix: tables of units
Table 1 SI base unils
Unit
Quantity Name Symbol
Length meter m
Mass kilogram kg
Time second ,
Electric curreOl amp<'C A
Temperature keh'in K
Amount of substance mole mol
Table 2 51 derived UMs
Unit
Quantity Name Symbol Equivalent
Plane angle radian rod nt/m = I

Solid angle steradian m21m 2 = I
Speed, velocity " m S~I
Acceleration m S-l
Angular velocity md S-I
Angular acceleration rod S-2
, I
Frequency hertz Hz ,
Force newton N kg m s -
Pressure. stress pascal Pa m-'
Work. energy. heat joule J m. kg m 2 S-l
Impulse. momentum s.kgms- t
Power \wlt W J s -I
Electric charge coulomb C A,
Electric potential. emf volt V JC-I • WA- I
Resistancc ohm n V A-I
Conductance siemens S A V-I. 0- 1
Magnetic nux weber Wb V,
Inductance henry H WbA- 1
I
Capacitance farad F CV
Electric field strt:nb'1h Vm- I • NC- I
Magnetic flux density tesla T \Vb m- 2. N A-I m- I
Electric displocement Cm- 1
Magnetic field strt:ngth Am-I
Celsius temperalure degree Celsius C K
Luminous flux
Illuminance
lumen
lux
1m
Ix
cd sr
Imm
,
Radioactivity bccquerd Bq ,-1
Catalytic activity katal kal mol S-I
324 Appendix
Table 3 Physical Constants
Constant Symbol Value Units
Speed of light in vacuum c 2.9979 x 108 III S-I
::::: 3 x 108
Elementary charge e 1.602 x 10- 19 C
Electron mass /110 9.11 x 10-ll kg
Electron charge to mass ratio e/lllo I. 76 X lOll C kg- l
Proton mass /lip 1.67 x 10- 27 kg
Boltzmann constant k, 1.38 x 10- 23 J K- 1
Gravitation constant G 6.67 x 10- 11 m) kg- I S-2
Standard acceleration of gravity g 9.807 m s-2

Permiuivity of free space 8.854 x 10- 12 F m-I
" '" W"/(36~)
Permeability of free space
Planck's constant
",h 4n x 10-7
6.6256 x I o-~
H m- I
Js
Impedance of free space '10 = JJ1.o/ff.O 376.73 ::::: 120JT Q
Avogadro constant NA 6.022 x 1023 mo!-I
Table 4 Standard prefixes used with SI units
Preftx Abbreviation Meaning Preftx Abbreviation Meaning
atto- a- J 0- 18 deb- da- 10'

femto- r- la-Is hecto~ h- 102
PICO- p- 10- 12 kilo- k- 10'
nano- n- 10-9 mega- M- 10'
mlcro- ~- 10- 6 glga· G- 10'
milti· m- 10-3 tera· T- 10 12
ccnti- c- 10- 2 pcta- p- lOIS
deci- d- 10- 1 exa- E- 10 18
Table 5 Conversion of SI units 10 Guassian units
Quantity Si unit Gaussian units
Length 1m 102 em
Mass I kg 10l g
Force IN lOS dyne = lOs gem S-2
Energy IJ 107 erg = 107 g cm 2 s-2
I eV = I.602 x 10- 19 J = \.602 x 10- 12 erg

#Mitin - Introduction To Nano Electronics

Uploaded by

Copyright:

Available Formats

#Mitin - Introduction To Nano Electronics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

#Mitin - Introduction To Nano Electronics

Uploaded by

Copyright:

Available Formats

Introduction to Nanoelectronics

Science, Nanotechnology, Engineering,

Increasing miniaturization of devices, components, and integrated systems requires

Vladimir V. Mitin is a Professor and Chair of the Department of Electrical Engineering

Viatcheslav A. Kochelap is Professor and Head of the Theorctical Physics Departmcnt

Michael A. Stroscio is a Professor in the departments of Electrical and Computer Engi-

Cambridge University Press

© Cambridge University Press 2008

This publication is in copyrighr. Subject to Sr3[U[ory exception and (0 the provision of

First published in prill{ [ormar 2007

ISBN-I3 978-0-51 1-36964-3 eBook (ebra'l)

ISBN-13 978-0-521-88172-2 hardback

including nanoscience, physics of material devices, electrical and optical engineering,

nl"FI~olc 'n''Of1~I~"''~ Jil

nl ' , a !~~Io.,,- 'n "~~.1 "~Ic,,~

Table 1.1 Aroadmap for Si-based microelectronics (predictions of the Semiconductor

1995 1998 2001 2004 2007 2010

350 '--,~.~•.,.._p_1_a_na_'_C_M_O_S_D_OV_;_'. 1 Novel Device ConCQpts

complexity, microelectronic systems continue to be produced at lower costs. In Table 1.1

Improving materials on lt1e nanoscale

materials. With modern material-growth techniques, it is possible to grow structures

Fabrication techniques on the nanoscate

Table 1.2 Advances in growth, characterizatioo. and processing of quantum

1970s-1980s Growth and fabrication methods

Improvement in characterization methods for the nanoscale

.. ...,- ... ,,, .....-

the electronic and lattice properties of heterostructures. Finally. terahertz time-domain

New principles of device operation at the nanoscale

Fundamental questions arise when conventional principles of device operation fail as

Nanotechnology for optoelectronics

R. Feynman, 'There's plenty of room at the bOllom," American Physical Society

2.2 Classical particles

2.3 Classical waves

11-2 11-1 11 11+1 11+2

In = -fJ(u n -Un+l) - fJ(u n - lin-I). (2.11)

By integrating the second term by parts by using the standard relationship J w dv =

:,~ t dz[PtC';j + A~DC~)']

To draw further conclusions, we define an average of a quantity, A, over a piece of the

U(1. r) = A cos(qr - WI) + B sin(qr - WI). (2.20)

U(I. r) = Bsin(qr - WI). (2.22)

= ~'(p~' +q:,,) = ~'(p~' +p;') = ~'pw'

0,(1. r) = B, sin(wl + I',(r». 0,(1. r) = B, sin(wl + I',(r)). (2.27)

In deriving Eq. (2.28), we have used the identity

Agure 2.3 A double-slit experiment: 51 and 51 form the double slit.

o(t. r) = B+ sin(qr - wt) + B_ sin(qr + wI).

(a) II = I (b) II = 2 (c) 11=3

u = B sin(qnz) COS(WIt/). (2.30)

U(/. r) = L Aqei(qr-wql). (2.31 )

Electromagnetic waves in free space

8'F2 _ c' (8 2 F2 + 8'F + 8'F) = O. (2.32)

F(r. I) = bF" sin(qr - wI).

F(r, t) = _ibFoeililr-wt), (2.33)

In free space, the vector F is always perpendicular to q, so that, if q is fixed, the

a'F c' (a'F a'F a'F)

2.4 Wave-particle duality

From waves to particles

where tl = 1.06 x 10- 34 J s (Joule-second) is the so-called reduced Planck COllstallt;

Table 2.1 Acomparison between classical and quantum quantities

Classical quantity Corresponding quantum quantity

nl ' , a !Io.,,- 'n ".1 "~Ic,,~