Functional Thin Films and Nanostructures For Sensors. Synthesis, Physics, and Applications, 2009, p.224

Functional Thin Films and
Nanostructures for Sensors

Integrated Analytical Systems
Series Editor: Dr. Radislav A. Potyrailo

GE Global Research, Niskayuna, NY
For other titles published in this series, go to

www.springer.com/series/7427
Anis Zribi • Jeffrey Fortin
Editors
Functional Thin Films and

Nanostructures for Sensors
Synthesis, Physics, and Applications
Editors
Anis Zribi Jeffrey Fortin
United Technologies Corporation GE Global Research Center
Fire and Security Micro and Nano Structures Technologies
Kidde Detection Technology Research Niskayuna, NY
Development and Engineering USA
Colorado Springs, CO
USA
ISBN: 978-0-387-36229-8 e-ISBN: 978-0-387-68609-7

DOI: 10.1007/978-0-387-68609-7
Library of Congress Control Number: 2008944096
© Springer Science+Business Media, LLC 2009

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they
are not identified as such, is not to be taken as an expression of opinion as to whether or not they are
subject to proprietary rights.
Printed on acid-free paper
springer.com
To our families and parents.
Olena, Nadia
Michelle, Abi, Libby
Foreword
In recent years, there has been a convergence of fundamental materials science and
materials processing methods. This convergence, although highly interdisciplinary
in nature, has been brought about by technologies such as bandgap engineering and
related techniques that have led to application-specific devices such as lab-on-a-
chip and system-on-a-chip. The demand for reduced device size, device portability,
and low power dissipation coupled with high speed of operation continues to dictate
terms and conditions for the evolution of nanotechnology. The present trend in
approaches to systems manufacturing continues to focus on integration of multi-
functionalities on the same chip. These functionalities include, for example, on-
board laser sources, sensors, and amplifiers. Both the military and civilian markets
continue to drive the research and development component. In recent years, the
emergency preparedness guidance systems have added excitement and curiosity to
this expanding industry. The outgrowth of technologies of interest for emergency
preparedness includes the development of terahertz sources and detectors and sys-
tems for detection of explosives and concealed weapons, among others.
Sensors made from bulk materials have been around for a long time. Enormous
advances in the processing technologies of thin films have led to the ability to
manufacture application-specific functional thin films. These include transparent
electrodes and antireflection films such as indium tin oxide, which serve as inter-
face components between humans and electronic devices, or optical circuit ele-
ments used in optical communication networks, or as contacts and antireflection
coatings in solar cells. Products are also being developed with magneto-optical,
electrochromic, or UV material for their use as functional thin films in optics.
Photonic crystals contain a variety of functional thin films; they require processing
of thin films under very stringent control of their structure and properties.
For microelectromechanical systems (MEMS), in addition to silicon-based tech-
nology, ferroelectric thin films are being used in the fabrication of microactuators
and micromotors, capacitors, and other thin-film devices. Functional thin films are
being used in the manufacture of devices such as surface acoustic wave (SAW)
devices for high-frequency telecommunications filtering, infrared detectors, pres-
sure sensors, accelerometers, force sensors, vibration, thickness, and chemical sen-
sors and biosensors. The reduction in size from bulk to micro- and nanostructured
transducers, while promising high sensitivity, high speed, and increased selectivity,
vii
viii Foreword
requires new design considerations that should consider factors such as integration
with other devices and device lifetime.
Functional thin films offer an enormous infrastructure for a highly interdisciplin-
ary integration of inorganic/semiconducting, organic/bio, and electronic/optoelec-
tronic sensor systems. The field is constantly evolving and will continue to do so
by absorbing novel materials approaches such as carbon nanotubes, high Tc super-
conductors, ferroelectrics, and thermoelectrics.
The chapters in this book are designed to give the reader the big picture, from
the design phase to the implementation and realization of a transducer. Every effort
has been made to include the state-of-the-art in each chapter. The intended audience
is scientists, researchers, and engineers, however, graduate students will find the
book to be very useful in their research and understanding of sensors and beyond.
The editors and contributors are leading researchers in industry and academia in
their subject areas.
Newark, New Jersey N. M. Ravindra
February 2008
Series Preface
In my career I’ve found that “thinking outside the box” works

better if I know what’s “inside the box.”
Dave Grusin, composer and jazz musician
Different people think in different time frames: scientists think

in decades, engineers think in years, and investors think in
quarters.
Stan Williams, Director of Quantum Science
Research,Hewlett Packard Laboratories
Everything can be made smaller, never mind physics;

Everything can be made more efficient, never mind
thermodynamics;
Everything will be more expensive, never mind common sense.
Tomas Hirschfeld, pioneer of industrial spectroscopy
Integrated Analytical Systems
Series Editor: Dr. Radislav A. Potyrailo, GE Global Research, Niskayuna, NY
The book series Integrated Analytical Systems offers the most recent advances in all
key aspects of development and applications of modern instrumentation for chemi-
cal and biological analysis. The key development aspects include: (i) innovations in
sample introduction through micro- and nanofluidic designs; (ii) new types and
methods of fabrication of physical transducers and ion detectors; (iii) materials for
sensors that became available due to the breakthroughs in biology, combinatorial
materials science, and nanotechnology; and (iv) innovative data processing and
mining methodologies that provide dramatically reduced rates of false alarms.
A multidisciplinary effort is required to design and build instruments with previ-
ously unavailable capabilities for demanding new applications. Instruments with
more sensitivity are required today to analyze ultratrace levels of environmental
pollutants, pathogens in water, and low vapor pressure energetic materials in air.
ix
x Series Preface
Sensor systems with faster response times are desired to monitor transient in vivo
events and bedside patients. More selective instruments are sought to analyze
specific proteins in vitro and analyze ambient urban or battlefield air. For these and
many other applications, new analytical instrumentation is urgently needed. This
book series is intended to be a primary source of both fundamental and practical
information of where analytical instrumentation technologies are now and where
they are headed in the future.
Looking back over peer-reviewed technical articles from several decades ago,
one notices that the overwhelming majority of publications on chemical analysis
has been related to chemical and biological sensors and has originated from depart-
ments of chemistry in universities and divisions of life sciences of governmental
laboratories. Since then, the number of disciplines has dramatically increased
because of the ever-expanding needs for miniaturization (e.g., for in vivo cell analy-
sis, embedding into soldier uniforms), lower power consumption (e.g., harvested
power), and the ability to operate in complex environments (e.g., whole blood,
industrial water, or battlefield air) for more selective, sensitive, and rapid determi-
nation of chemical and biological species. Compact analytical systems that have a
sensor as one of the system components are becoming more important than indi-
vidual sensors. Thus, in addition to traditional sensor approaches, a variety of new
themes has been introduced to achieve an attractive goal of analyzing chemical and
biological species on the micro- and nanoscale.
Preface
Anyone with the most cursory knowledge of sensors must have had a chance to use
such devices at some point in their life or career. Whether to collect data in a lab
course, to automate an otherwise tedious process, to improve the efficiency of a
delicately tuned process, or to do something as mundane as taking a family picture,
sensors have become an integral part of our environment and our daily lives.
Charge-coupled devices also known as CCD photodetector arrays, for example,
have revolutionized photography, astronomy, spectroscopy, X-ray diffraction, and
medical imaging to name but a few. A number of scientific discoveries have been
enabled by CCDs including the possibility to determine molecular and lattice struc-
tures at intermediate stages of a chemical synthesis or a structural transformation.
At the core of the widespread adoption of sensors are their rapidly decreasing foot-
print and cost and increased functionality. The miniaturization of solid-state devices
in general and sensors in particular was made possible thanks to significant trans-
formations and a large number of incremental and disruptive inventions in the area
of thin-film and nanostructure science and fabrication technologies.
Thin films and nanostructures can play multiple roles in a sensor including
structural support, reliability enhancement, filtering, and transduction. Thin films
and nanostructures are called functional when they fulfill a function other than
structural support. These micro- and nanostructured materials have applications
that extend far beyond sensing to data storage, lighting, displays, hydrophobic coat-
ings, decoration, and a large number of other fields that are outside the scope of this
book. In this book, these materials are discussed in the context of transduction and
how they contributed to the current sensor revolution.
Sensor design and fabrication are multidisciplinary and require broad and deep
knowledge in diverse areas of science and engineering such as materials science,
physics, chemistry, biology, and mechanical and electrical engineering. Covering a
subject with so many roots in diverse scientific and engineering disciplines is
undoubtedly a daunting task and any author who attempts it will do so with signifi-
cant trepidation. Aware of the challenge at hand, the editors of this book attempted,
ambitiously, to cover in one volume an account of general sensor theory, design
considerations related to the use of functional thin films and nanostructures, and
specific case studies of functional thin films and nanostructure applications in
sensing. Part of our motivation in taking on this task is that no such work, to our
xi
xii Preface
knowledge, has been published. Having said this, we are strongly familiar with the
large body of publications in this area that we refer to in this book and we are
keenly indebted to the works of many authors in putting this book together.
This book is devoted to teaching the new sensor designer the key steps involved
in developing sound transducer technology from materials selection, to design for
performance, to process development, and finally to integration. Throughout the
chapters, the authors emphasize and highlight the important role played by func-
tional thin films in solving problems and discuss how to take advantage of such
materials to build superior devices. The book is also intended to provide the more
experienced designers with a condensed summary of sensor design methodology
and excellent references that will prove useful in future sensor design endeavors.
To put all of the shared design and fabrication knowledge into perspective and add
a touch of reality to the concepts discussed in Chapters 1 through 4, Chapters 5
through 8 are completely dedicated to putting the theory into practice and demon-
strating the whole design process using a number of concrete applications.
February 2008 Anis Zribi
Jeffrey Fortin
Editor Biographies
Anis Zribi is the manager of the Detection Technology Research, Development

and Engineering group at Kidde UTC Fire and Security. Prior to joining UTC, he
was a senior scientist and a principal investigator at the Global Research Center
(GRC) of General Electric where he (leads) led research in the area of Microsystems
and microfluidics for chemical and biological detection. He received an M.S.E.
in physics from the Polytechnic Institute of Engineering (1996 France), an M.S. in
materials physics from Chalmers University of Technology (1998 Sweden) and a
Ph.D. in materials science from the State University of New York (2002 NY). Since
joining GRC in 2002, Dr. Zribi has contributed to and led several projects including
the Nanotechnology Advanced Technology Program, the Photonics Advanced
Technology Program, and a number of MEMS sensors and actuators projects. His
research interests and activities at GRC include MEMS spectrometers, chemical
and biological sensing, magnetic field sensing, medical parameters sensing, fouling
detection, and micro- and nanotransducers. Dr. Zribi holds 7 patents and over 40
pending patent applications in MEMS, photonics, and sensors. He authored or
co-authored more than 32 articles in peer-reviewed journals and conference pro-
ceedings and two book chapters.
Jeff Fortin is the manager of the Microsystems and Microfluidics Lab at GE

Global Research in Niskayuna, NY. His team’s charter is to develop and deliver
innovative micro- and nanosystems and microfluidics via the development and
integration of MEMS and NEMS sensing, actuation, and microfluidic technologies,
driving miniaturization, increased performance, portability, and low cost. Jeff holds
a Ph.D. in engineering science from Rensselaer Polytechnic Institute, an M.S. in
physics from RPI, and a B.A. in physics from the University of Southern Maine.
He has over ten years of experience in semiconductor technology, MEMS, and
microsensors. He joined GE GRC in 2000 and since this time his research has
focused on MEMS and microsystems design and fabrication for a variety of
microsensor and microactuator applications for GE. He holds ten patents and has
co-authored over 12 refereed journal articles in the area of MEMS, thin polymer
film development, and chemical vapor deposition as well as eight conference pub-
lications. He is also the co-author of a text on chemical vapor deposition polymer-
ization of parylene and is a member of the MRS.
xiii
Contents
1 Sensor Design Guidelines ........................................................................ 1

Anis Zribi
2 Transduction Principles........................................................................... 17
Jeffrey Fortin
3 Growth and Synthesis of Nanostructured Thin Films ......................... 31

Yiping Zhao
4 Integrated Micromachining Technologies

for Transducer Fabrication ..................................................................... 65
Wei Cheng Tian
5 Applications of Functional Thin Films

and Nanostructures in Gas Sensing ....................................................... 85
Audrey Nelson
6 Chemical Sensors: New Ideas for the Mature Field ............................. 103
Radislav A. Potyrailo
7 Applications of Functional Thin Films for Mechanical Sensing ......... 145

Chang Liu
8 Sensing Infrared and Terahertz Regions by Functional Films ........... 167

Magnus Willander, Victor Ryzhii, and Qingxiang Zhao
Index ................................................................................................................ 211
xv
Contributors
Jeffrey Fortin
GE Global Research Center
Micro and Nano Structures Technologies
Chang Liu
Northwestern University
Department of Mechanical Engineering
Audry Nelson
GE Sensing, TelAire
Radislav Potyrailo
Vicktor Ryzhii
University of Aizu
Computer Solid State Physics Laboratory
Wei-Cheng Tian
Micro and Nano Structures Technologies
Magnus Willander
Linköping University
Institute of Science and Technology
Qingxiang Zhao
Linköping University
Institute of Science and Technology
Yiping Zhao
University of Georgia
Department of Physics and Astronomy
Anis Zribi
United Technologies Corporation Fire and Security, Kidde Detection
Technology Research, Development and Engineering Colorado Springs,
CO, USA
xvii
Chapter 1
Sensor Design Guidelines
Anis Zribi
Abstract This chapter focuses on introducing fundamental design principles of

transducers, familiarizing readers who are new to this field with the common vocab-
ulary used in describing transducer performance, and providing a succinct historical
background about the implementation of thin films and nanostructures in sensors
and analytical instruments. A systematic methodology and a sequence of guiding
steps to follow in designing a transducer beginning with a concept, through materials
selection, and transducer design and fabrication are presented. These steps are covered
in more detail in subsequent chapters with concrete examples.
The Big Picture
Sensors are ubiquitous in our environment and play essential roles in our everyday
life. Our own view of the world is defined by our senses that enable us to perceive
stimuli from the environment through a network of biological sensors. Tiny hairs in
our inner ears detect the deflection of a membrane as it vibrates in response to
acoustic waves and make it possible for us to hear; photoreceptors in our eyes
enable us to see objects and discern their colors; chemical receptors on the tongue
(known as taste buds) allow us to differentiate between salty, sweet, bitter, and sour.
This fascinating network of biological sensors caters to our organs’ needs to control
certain biological processes and our needs for security and safety.
Driven by the need to better understand our world, to increase the productivity
of industrial processes and machines, and to improve our quality of life, scientists
and engineers constantly seek to develop the necessary measurement tools. These
tools are sensors and instruments that are often inspired by biological sensors and
their functioning principles. Such devices have become essential for advancing
A. Zribi
United Technologies Corporation Fire and Security, Kidde Detection Technology Research,
Development and Engineering 4820 Centennial Blvd, Suite 145, Colorado Springs 80919,
CO, USA
e-mail: anis.zribi@fs.utc.com
A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, 1
Integrated Analytical Systems,
DOI: 10.1007/978-0-387-68609-7_1, © Springer Science + Business Media, LLC 2009
2 A. Zribi
metrology and science, optimizing and controlling processes, providing security

against known and unknown threats, law enforcement, and health monitoring.
The miniaturization of sensors and analytical instruments is a continuing trend
that finds its roots in nature and its beginnings in miniaturizing mechanical, optical,
and recently electronic devices. The driving forces for sensor miniaturization are
numerous and keep increasing as we encounter and develop new applications for
devices that have become virtually invisible and intangible. Sensor cost and porta-
bility are major incentives for miniaturization but additional reasons include faster
response, integration of multiple functions, reduction of occupied space, and lower
device-to-device variability.
Electronic miniaturization, starting with the invention of the transistor in 1947,
played a key role in developing the fabrication processes that stand at the heart of the
new leap in sensor miniaturization. In fact tremendous progress has been made in the
past 20 years in high vacuum technologies, ultra purification processes for raw mate-
rials, self-assembly, and material deposition and etching processes with various
degrees of selectivity, high precision, and small feature-patterning techniques. This
progress produced dramatic advances and control over device quality and induced a
technological evolution from the bulk crystalline age to the age of thin films, thin film
multilayers, and nanostructured materials. In this new age, the properties and perform-
ance of submicron devices are dominated by surface and interfacial phenomena. This
paradigm shift produced materials with novel or enhanced transduction properties
known as functional thin films and functional nanostructures.
Thin films and nanostructures play an increasingly important role in state-of-the-art
sensors and actuator technologies both as transducers (functional materials) and struc-
tural materials. microelectro mechanical (MEM) systems provide a good example of
the growing use of materials confined to submicron dimensions to fulfill numerous and
versatile functions in advanced devices. Doped silicon/polysilicon thin films, for
example, have been implemented as strain gauges in MEM pressure sensor devices
(Petersen 1982), temperature-sensing elements in micromachined thermopiles
(Petersen 1982), ultrasound transmitters in capacitive ultrasound micromachined
transceivers (cMUT; Jin et al. 1998), active alignment actuators in high-accuracy fiber-
optic aligners (Petersen 1982), and the list goes on. The growing interest in functional
thin films and nanostructures is not only driven by device miniaturization but also by
the novel and unique set of physical and chemical properties that materials confined to
submicron dimensions exhibit.
In his famous talk before the audience of the 1959 annual meeting of the
American Physical Society, Richard Feynman predicted many of the now proven
advantages of “manipulating and controlling things on a small scale” (Feynman
1959). Since Feynman’s talk, the advantages of scaling devices and materials to
submicronic dimensions have been proven to go beyond a dramatic increase in data
storage density, faster and more intelligent computing, faster heat removal, and
higher natural resonance frequency.
Today, functional thin films and nanostructures are being used and are under
development for use in a wide range of devices, sensors, and actuators. Such devices
include: accelerometers (air bag devices), force sensors, shutters, optical switches,
1 Sensor Design Guidelines 3
optical computation, micromotors, chemical and biological sensors; MEMS devices

such as piezomicroactuators and sensors; very small scale microreaction vessels for
chemical and biological (lab-on-a-chip) sensing; substrates for plasmon resonance-
based signal enhancement and amplification, electro-optical devices for thermal
imaging based on the pyroelectric effect, display technologies; pressure mats (piezo-
electric thin film); surface acoustic wave (SAW) devices for telecommunications,
chemical and biological detection, and more.
These new applications for functional thin films and nanostructures are enabled by
newly discovered physical and chemical phenomena that underlie the transduction
mechanisms in these materials. In fact, size confinement and the associated symmetry
breaking, interfacial interactions, high surface-to-volume ratio, structural disorder, and
induced entropy have been extensively covered in recent publications. The fundamen-
tals of these effects and their applications in transduction are the topic of numerous
current investigations. Quantum dots, for example, have been demonstrated to exhibit
a lateral quantum confinement that enables direct coupling of normally incident light
with the intraband electronic excitations (Kuo et al. 2001). The geometry of these
nanostructures (height and radius) can be tuned to ensure that the lowest energy of
electronic transition from ground state to first excited state falls within a specific optical
spectrum band. It has also been demonstrated that the confinement significantly
reduces the electronic tunneling rate of these nanostructures. These properties indicate
the potential of quantum dots to be used as low dark current and hence high signal-to-
noise ratio photodetectors especially in the near infrared to infrared part of the spec-
trum where current detectors are prone to thermally induced electronic tunneling. This
and other transduction benefits that emanate from the physical confinement of material
structures are discussed in later chapters of this book.
Sensor Architecture
The basic function of a sensor is to selectively identify and measure a physical,

chemical, or biological parameter such as pressure, light intensity, gas concentration,
or the presence and concentration of a biological analyte. The typical architecture of
a sensor encompasses a transducer or multiple transducers (operating in series or in
parallel) directly exposed to the measurand, acquisition and conditioning electronics,
a power source, a processor, a storage medium, and a display. The transducer plays a
central role in the operation of the sensor: it is essentially an energy converter where
the input energy (mechanical, optical, chemical, biological, electrical) is converted
into an electrical signal most of the time. The electrical signal is then acquired by the
electronics, conditioned, and noise filtered out before processing (e.g., interpolation
using a calibration curve) and finally the data (typically the magnitude of the meas-
urand) are either stored in memory and displayed or routed for action (e.g., alarm) or
simply displayed.
Fig. 1.1 shows a general block diagram that highlights the main subsystems of
a sensor system and the flow of data among the various blocks. The focus of this
4 A. Zribi
Measurand M Sensing X Y Y1 Y2 Numerical

Transducer Amplifier Filter A/D Value N
Material
Power Block
Processor Storage Display
Fig. 1.1 Sensor architecture block diagram
book is on the transducer block and the role that functional thin films and nanos-
tructures are increasingly playing in high-performance transduction.
Depending on the type of transducer, two basic types of sensors can be distin-
guished: quantitative (analog or digital) and threshold or binary. The two are quite
different in function and in application. A quantitative sensor produces an output
value that is a direct and continuous function of a measurand value. For example, a
thermocouple might have a potential differential of 10 mV at room temperature and
a potential differential of 20 mV 10° above room temperature. Any differential
potential value between these two is possible depending on the particular tempera-
ture to which the sensor is exposed. Threshold sensors, on the other hand, have only
two states, often called “on” and “off”. Perhaps the most familiar example of a
threshold sensor is a smoke detector which is triggered to the on position if a fire
erupts and the signal is used to trip an alarm.
Sensor Figures of Merit/Performance Attributes
The wealth of sensor and transducer technologies available to measure the same
measurands (Mi) increases the challenge of evaluating and comparing the perform-
ance of sensor devices. Therefore, it is critical to define a set of performance criteria
that the designers can use to develop sensors that meet customer specifications and the
user can use to appraise and contrast the various options. Table 1.1 summarizes the
list of performance attributes that are most commonly used to assess sensor tech-
nologies and a more detailed description of these parameters is provided later in
this chapter.
Table 1.1 Sensor Performance Attributes

Attribute Denotation
Input dynamic range DM
Response curve N = f(M)
Sensitivity S
Response time t
Resolution n
Accuracy
Precision
Hysteresis and drift H
Selectivity Se
N
Mimin Mimax
Mi
ng
reasi
Inc
Mij
δN sing M i
Decrea
δMi
Mi
Input Dynamic Range
Fig. 1.2 Typical response curve of a sensor/transducer to a measurand Mi
Input Dynamic Range
The dynamic range of a sensor is the span of measurands that constitute the overall
operating domain for the device. Within this interval, the sensor is supposed to
maintain its properties and reliability characteristics.
Response Curve
Every quantitative sensor is characterized by a response curve that represents the

output of the sensor N versus the measurand M applied to its input. The transducer
response can be linear or nonlinear as shown in Fig. 1.2, but in most cases the sen-
sor electronics are designed to linearize the response curve of the overall sensor in
6 A. Zribi
order to simplify the calibration procedures. Also, it is worth noting that customar-
ily the response of the sensor is normalized with respect to reference values (e.g.,
resistance at room temperature, resonance frequency of the membrane under known
conditions).
Considering all the possible measurands (Mi) and key noise parameters that can
affect the sensor output, the sensor response N can be expressed as
∂N
N = ∑ ΔMi (1.1)
i ∂M i
given that the different measurands affect the output independently of each other.
Sensitivity
Although seemingly obvious, the definition of sensitivity can be confusing at times.

Looking at the block diagram of the sensor in the first section, each one of the
building blocks of the sensor introduces a term in the overall sensitivity of the
sensor. These are called, following the diagram from input to output:
• Transducer sensitivity
• Amplification
• Analog filter sensitivity
The sensitivity of transducer j to measurand Mi is defined by
∂N j
j =
i
Stransducer , (1.2)
∂M i
where Mi are all the possible measurands and key noise parameters that can affect
the transducer sensitivity such as temperature and pressure. Nj is a quantity resulting
from the energy transformation performed by transducer j. Transducers in a detector
can operate in parallel or in series. Typically, transducers that operate in parallel are
used for imaging applications, which we are not concerned with within this manuscript.
Assuming the p transducers that make up the transducer block operate in series, the
total transduction block sensitivity to measurand Mi can be expressed by
∂N p ∂N p ∂N p − 2 ∂N1 p
i
Stransducer = ... = Π S kj . (1.3)
∂M i ∂ M p −1 ∂ M p −1 ∂M i k = 1
The amplification A is an intrinsic property of the amplifier and the analog filter
sensitivity is defined in a similar manner to the transducer block by Equation (1.3):
∂Y2 (1.4)
Sfilter = .
∂Y1
The overall sensitivity of the device to measurand Mi is defined by the following

equations,
∂Y2
Si total = (1.5)
∂M i
∂Y2 ∂Y1 ∂N
Si total = . . (1.6)
∂Y1 ∂N ∂Mi
Si total = Sfilter * A * Stransducer . (1.7)
Response Time
The response time of the device is a cumulative quantity incorporating the response
times of the respective building blocks. It is defined as the transient response of the
sensor when it experiences a step change in the measurand. If ti are the response
times of the various blocks, then the total response time of the sensor is:
t total = t transducer + t filter + t A / D (1.8)
t total = t transducer + t electronics . (1.9)
In most diffusion-based chemical and biological sensors, tchemical transducer is the domi-
nant response time in Equations (1.7) and (1.8). Therefore, we typically can ignore
the contributions from electronics and physical transducers. However, this may not
be the case in thin-film or nanostructure-enabled sensors as the response time of the
material is considerably reduced because of the designed nanomorphology of the
material. This simplification is also not an option for physical sensors where the
response time of the transducer is on the same order of magnitude as the other sen-
sor subsystems.
The response time of the whole device is usually estimated as the time required for
a transient output signal to reach a fraction (e.g., 70%) of its steady-state change.
Resolution
The sensor resolution is the smallest change in input that leads to a detectable
change in output. The most general definition is given by Equation (1.10):
N N (1.10)
Re solution = Lim = noi
M i → noise Si total Si tot
8 A. Zribi
Clearly for nonlinear sensors, the resolution is dependent on the operating point and
can vary if operating conditions vary. Maximum resolution is attained at maximum
sensitivity and minimum noise.
Accuracy
The accuracy of a transducer is the maximum deviation of its output from the value
of the unknown measurand as determined by a gold standard technique. It is a
quantitative indication of the degree of conformity of the sensor to a standard.
Accuracy represents a systematic error or a bias in the sensor and measuring it and
correcting for it are difficult. Transducer calibration is required to account for this
type of error and partially correct it. Long-term changes in the performance of the
transducer because of materials’ aging and wear will affect its accuracy.
Precision
The second type of measurement uncertainty associated with a transducer is

random errors or what is often referred to as precision. Precision is often mistakenly
confounded with accuracy. It quantifies statistical fluctuations in the measurement
and is attributed to variability in the measurement conditions and the limitations of
the selectivity of the sensor towards the measurand of interest. Precision represents
the repeatability of the measurement given the same sample.
Hysteresis and Drift
Hysteresis characterizes the lagging of the sensor response behind the variation of the
measurand. This can be attributed to the sensing material memory and/or to the trans-
ducer properties. As a performance specification, hysteresis is defined as the maximum
difference between the upscale and downscale readings on the same artifact during a
full-range traverse in each direction. It is often reported as the ratio (or percentage) of
the difference between the upscale and downscale readings to the full scale.
Drift can be defined as the slow unpredictable change of the sensor output at
constant input. Drift can affect both the signal and noise levels and it can emanate,
for example, from residual stress relaxation, residual diffusion, material aging, and
degradation. Drift in sensors is defined for a specific time interval of interest.
Selectivity
It is typical of a sensor designed to detect variations of a given measurand M to

exhibit secondary internal sensitivities to other measurands (Mi) or noise factors of
different natures, physical or chemical, as well as biological. In general, selectivity

may be quantified by
p
Stjransducer
Se = ∏ ,
j =1 St1ransducer (1.11)
j ≠1
where Sj is the sensitivity of transducer j to measurand Mi. In many cases, the

sensitivities to the various measurands are interdependent therefore second-order
and cross-terms should be included in the expression of the selectivity.
Sensor Design Considerations
Sensor design is one of the most interdisciplinary technical areas requiring both
breadth and depth of knowledge in materials science, physics, chemistry, biology,
mathematics and statistics, electronics, and packaging. The successful design of a
sensing system requires very good communication between scientists and engi-
neers of different backgrounds.
Much as in any design activity, it is very difficult and undesirable to bind creativity
by a set of design rules or a design methodology. However, past a first stage, focused
on brainstorming, transduction mechanism and materials down-selection and feasi-
bility analysis, a successful design team must have a clear objective and a guiding
design methodology to steer their effort. The typical sensor design steps include:
• Collection and analyses of the device specifications
• Transducer selection/invention
• Materials selection/invention
• Sensor design and modeling
• Prototyping
• Measurement of materials properties
• Prototype testing and model calibration
• Design iteration(s)
• Final device fabrication
• Technology transition to manufacturing
• Fabrication process scaleup
The most challenging and least regulated steps are the first three and they are least
covered by the literature. In the next sections, we provide some guidance, concepts,
and ideas to help the new sensor designer make faster progress towards his or her
ultimate goal.
Selection and/or Invention of the Transduction Mechanism
Assuming that the design team is armed with a clear set of specifications for the
desired sensor performance and the operating and storage conditions, the next steps
10 A. Zribi
ought to target the selection or invention of the transducer(s) that can convert the
measurand into an electrical/optical signal. The questions that need to be answered
by the team at this stage are:
1. What specific physical, chemical, and biological properties does the measurand
possess that differentiate it from other potential sensor inputs?
2. For each of the identified characteristics, what are the potential confounding
inputs, which in the future will be identified as noise sources?
3. What type of energy conversion is required to transform the input signal (mechani-
cal, chemical, biological, electromagnetic) to a device output signal (optical or
electrical)?
4. What are the candidate transduction schemes that can be used for the target
measurand?
The final transduction scheme may involve one or multiple transducers operating
in series or parallel (array of detectors). Considering the following criteria can
further refine the list of candidate transduction mechanisms.
1. Simplicity of the transducer and robustness to failure
2. Materials requirements dictated by operating environment
3. Development time
4. Cost
Selection of Transducer Material
The transducer material properties are important from performance and sensor reli-
ability perspectives. The suite of functional materials available to sensor and
micro-/nanoinstrument designers is rapidly expanding and numerous techniques
have been developed to integrate a large variety of organic and inorganic materials
and their alloys. The wealth of options puts the designer face to face with the chal-
lenge of selecting the best-suited materials given a transducer design concept.
Numerous considerations come into play when selecting a transducer material
and they fall into three categories: performance, process compatibility, and reliability.
Although it is very difficult to discuss the performance selection criteria of a trans-
ducer in general terms, process compatibility and reliability criteria are common to
all transducers and they are of chemical (chemical resistance, photo definability,
adhesion) and physical nature (rheology, mechanical, dielectric). Depending on the
transducer length scale, whether it falls in the bulk (>100 mm), micro (<100 mm),
or nano range (<100 nm), materials properties and their behaviors and responses to
environmental conditions and stimuli are significantly different.
Over the years, bulk materials properties have been well documented and often
standardized, however, submicron materials properties are still not as well docu-
mented and many are being investigated (Srikar and Spearing 2003). Some of the
challenges associated with the measurement of micro- and nanomaterials properties

are related to the lack of analytical instruments for this regime but many are related
to the high anisotropy of materials properties and to the strong dependence of these
properties on the sample geometry and synthetic technique used to prepare the
sample. In certain cases, it will prove essential to perform materials characteriza-
tion to provide a sound sensor system design; in other cases initial guesstimates of
the materials properties using bulk materials data and physicochemical laws com-
bined with system-level testing, failure analyses, and design iterations will be suf-
ficient to develop a robust sensor.
The following section is by no means a rigorous treatment of physical and
chemical confinement effects on materials. The objective is to give the reader a
flavor of some of the expected effects of size confinement on various materials
properties and provide references where further details can be found.
Physical and Chemical Considerations
As device dimensions scale down to hundreds of microns and single digit microns,
structural and functional materials’ dimensions that make up the device are
approaching the size of a few atomic layers and molecules, a range now familiar to
most of us as the nanoregime (<100 nm). In this nanoworld, the surface-to-volume
ratio is extremely high and diverges as dimensions shrink down. In this dimensional
range, surface effects (curvature, surface energy) and structural defects dominate
materials’ properties. This can be further explained by the fact that intermolecular
forces such as Van Der Waals, London dispersion forces, ionic interactions, hydro-
gen bonds, and dipole–dipole interactions prevail at the nanoscale. In liquids con-
fined to nanodimensions, short-range order has been already observed and reported
in numerous studies and liquid thin-film thickness begins to take discrete values.
The implication is that the physical and chemical properties of nanomaterials
differ greatly from their bulk counterparts and they are often strong functions of
thermal fluctuations. These properties include the effective viscosity, diffusion
coefficients, melting point, glass transition temperature, refractive index, mechanical
properties (elastic modulus), and the thermal conductivity.
Melting Point
A number of studies established that the melting point of a geometrically confined

material is different from its bulk melting point (Alcoutlabi and McKenna1 2005).
Studies of the melting temperature (Tm) of organic, metallic, and ceramic nanoparticles
and thin films revealed that Tm decreases with the crystal size. The melting point
12 A. Zribi
depression of a particle of diameter d follows the Gibbs–Thomson equation

(Alcoutlabi and McKenna1 2005):
4s sl Tmbulk
ΔTm = Tmbulk − Tm (d ) = , (1.12)
dΔH f rs
where Tmbulk is the bulk melting point, Tm(d) is the melting point of a cylindrical
particle of diameter d, ssl is the surface tension of the solid–liquid interface, ΔHf
is the bulk enthalpy of fusion, and rs is the density of the solid.
This relationship is only applicable if certain conditions related to the isotropy

of the surface tension and invariability of the bulk enthalpy of fusion and density
are satisfied. These conditions are often violated in the nanoscale but the Gibbs–
Thomson relationship is still a good approximation of the melting point depression
of nanomaterials and will give a close estimate of the real melting temperature of
thin films and nanostructures. More complex melting mechanisms have been and
are still being developed and the reader is encouraged to consult this existing large
body of work to gain more insight into these models (Zhang M et al. 2000;
Coombes 1972; Lai et al. 1998; Zhang Z et al. 2000).
Glass Transition Temperature
The glass transition temperature is by definition the temperature below which

molecules have very little relative mobility. Different theories predict contradictory
effects of confinement on the glass transition temperature of partially or wholly
amorphous materials. Experimental measurements, however, indicate a glass tran-
sition depression accompanying the size reduction of materials (Alcoutlabi and
McKenna1 2005). Currently, there are no readily available theories to explain this
reduction and no formalism that enables the prediction of the confinement effects.
Elasticity and Plasticity
The presence of a higher fraction of atoms near surfaces and interfaces is charac-
teristic of thin films and nanostructures. The proximity of atoms to a surface or an
interface creates an atomic environment, different from the bulk, where surface free
energy plays a bigger role in the elastic and plastic behavior of materials. Film
thickness and grain size, for example, have been proven to affect the deformation
mechanisms in metallic films significantly (Lilleodden et al. 2001). Metallic films
with a thickness of a micron or less exhibit very different plastic behaviors depend-
ing on the grain size. Initially, films with larger grains exhibit pronounced hardness
whereas fine-grained films show a soft behavior. When exposed to a load producing
displacements on the same order of magnitude as the characteristic length scales of

the film or nanostructure, fine-grained films harden and large-grained films soften
due to strain–gradient plasticity. It has been established that the large number of
grain boundaries in fine-grained structures act as a continuous source of disloca-
tions responsible for the plastic behavior and the continuous load-displacement
behavior observed in these structures and absent from large-grained structures.
A theoretical treatment of the effect of surface free energy on the elastic modulus
of nanostructures can be found in Dingreville et al. (2004). In this publication, the
authors demonstrated that at nanoscales, the contribution of the surface energy to the
elastic modulus is not negligible anymore. The effective elastic bulk and shear
moduli of an isotropic spherical particle could be estimated using Equation (1.13):
3 ⎡ 2G 1 ⎤
Eˆ b = Eb + ⎢K − (9 L + 18 M + 8 N )⎥
4a ⎣ 3 Eb ⎦
(1.13)
1 ⎡1 2G ⎤
Eˆ shear = Eshear + ⎢ ( K + 6 m ) − 1 (3 M + 4 N )⎥ ,
a ⎣5 3 Eb ⎦
where Eb and Eshear are, respectively, the bulk and shear moduli of the material, K
and m are functions of the surface energy, Γ1 a first-order term in the power series
expansion of the surface energy density as a function of the surface strain, and L,
M, and N are the third-order elastic constants of the material. Equation (1.13)
clearly indicates that the contribution of the surface energy to the elastic modulus
is inversely proportional to the characteristic length of the structure. This term
increases the elastic modulus of materials in the nanoregime and is responsible for
the high stiffness of nanomaterials. An immediate consequence of high stiffness is
that nanostructures possess much higher mechanical resonance frequencies than
bulk structures. This property has been implemented by numerous investigators to
develop resonators for various applications including trace chemical and biological
detection (Calleja et al. 2005), high sensitivity pressure calibrators, and others.
Viscosity
Functional thin films and nanostructures can be solid or liquid phase. In the micro-
and nanosize regime, fluidic droplets and films differ from bulk fluids because of
their pronounced inhomogeneity. Micro- and nanofluids have been reported in the
literature (Pozhar 2000) to exhibit various rheological behaviors that are often
contradictory. These behaviors are based on fluidic models with scarce experimental
data and numerous assumptions regarding the nature and magnitude of interactions
of the fluid molecules with their surrounding.
One of the most accurate models of the viscosity of confined, inhomogeneous
molecular fluids has been formulated by Pozhar and Gubbins and thus named the
14 A. Zribi
PG model. The theoretical viscosity of an inhomogeneous fluid with the inhomo-

geneity in the z-direction can be estimated by using the PG expression shown in
Equation (1.14):
16 *
h( z ) = h[4p n* ( z )t * ( z )[1 + pb 0* ( z )]2 + p n ( z )b 0* ( z )]
5 (1.14)
where
5 m
h=( )
16s 2 pb
is the viscosity of a dilute hard sphere gas, b = 1 (k T ) , kB being the Boltzmann

B
constant, T the temperature, m the mass of a fluid molecule, and s the hardcore
diameter of the fluid molecule specific to the fluid–fluid hardcore intermolecular
interactions. n* (z) = n(z)s3 is the dimensionless equilibrium number density, n(z),
that of the nanofluid, and finally t * (z) is a dimensionless quantity that is propor-
tional to the viscoelastic relaxation time.
This model, described in detail in the literature (Pozhar 2000), predicts that the
average viscosity of a nanofluid confined to a volume that is a few times the
molecular diameter increases up to four times that of the bulk fluid for a given fluid
type, density, and temperature. The viscosity is dependent on the location within
the confining volume and approaches the average value as the critical dimensions
of the confining volume approach ten times the molecular diameter. Although not
yet supported by pertinent experimental data, this result is very valuable and is in
good agreement with measurements conducted on complicated fluids (Pozhar
2000). It constitutes a good approximation of fluid behaviors in the nanoscale and
can be used with caution to design transducers based on functional liquid-based
thin films or droplet.
Optical Properties
In addition to thermodynamic, mechanical, rheological, and electronic size effects,

a nanometer-range-confined material exhibits different optical properties from the
bulk counterpart. Numerous publications reported shifts in absorption bands
(Huang and Lue 1994), narrowing of absorption bands, surface plasmon resonance
(SPR), and attenuation of absorption bands (Truong and Courteau 1987) in metallic,
polymeric, ceramic, and composite materials at the nanoscale. These new effects
are attributed to quantum confinement of electrons and the resulting changes in the
electronic transitions between energy levels as well as collective conduction-band
electron plasma oscillations.
The optical properties of thin films and nanomaterials are not only challenging
to measure but also difficult to model. These properties are governed by the
dielectric function which consists of a real and an imaginary part. For structures
with dimensions larger than 10 nm, electrons still behave as do particles and the
classical size effect based on Drude’s model is applicable. For these length scales,
the dielectric function needs to be corrected for electron scattering. As the dimen-
sions of the thin film or particle near 10 nm, electrons behave more as waves and
Drude’s model is only applicable after introducing some energy range modifica-
tions that affect the dielectric function.
Various models (Wood 1982; Kawabata and Kubo 1966; Genzel et al. 1975;
Cocchini 1985; Bassani et al. 1985) built on different assumptions have been devised
to capture the absorption behavior of nanoparticles and thin films at a range of fre-
quencies with different degrees of success. Although many of these models predict
the correct trends for the absorption spectra, they often disagree numerically with
experimental data and diverge as the characteristic dimensions approach bulk length
scales (Huang and Lue 1994). In his publication, Huang derived an expression for
the real and imaginary parts of the dielectric function for small metallic particles
using Lindhard’s equation. Huang demonstrated with his model very good agree-
ment with experimental data for structures with sizes confined to less than 10 nm.
The indication is that absorption spectra of metallic structures in the nanoscale shift
towards the blue part of the spectrum and absorption peaks tend to broaden. These
findings can be accounted for and taken advantage of when designing optical com-
ponents with dimensions confined to the nanometer range. The optical properties of
metal nanoparticles and more specifically noble metal nanoparticles will remain a
continuous subject of research (Scarrafrdi et al., 2005, Scaffardi and Tocho, 2006)
because of the potential applications in many fields such as spectrally selective coat-
ings, nonlinear optics, and heterogeneous catalysis.
Summary
This introductory chapter summarized the general design guidelines of a sensor

regardless of the analyte the sensor is designed to measure, the environment the sen-
sor is designed to operate in, or the application. In real-life applications, all of these
parameters significantly affect the design from material selection to the selection of
the transduction scheme, the geometry of the device, and the package design.
It is well established that nano- and microstructured sensing materials provide
unique and novel functions unattainable using bulk materials. More specifically,
micro- and nanostructured transducers promise to be more sensitive, more selec-
tive, and faster responding, but this comes at a cost. At these scales, new design
considerations need to be taken into account including difficulty of fabrication,
integration with macroscale structures, sensitivity to environmental conditions,
stability (thermal and chemical), and long-term reliability. These issues are subjects
of numerous research efforts making great inroads towards bringing these materials
to mainstream everyday-life sensing devices. The next chapters delve into the vari-
ous aspects and details of implementing thin films and nanostructures into sensors
and the challenges and benefits of this endeavor.
16 A. Zribi
References
Alcoutlabi M, McKenna1 GB (2005) Effects of confinement on material behaviour at the nano-

metre size scale. J. Phys.: Condensed Matter, 17:R461–R524
Bassani F, Bourg M, Cocchini F (1985) Size effect in the optical properties of small metallic
particles: A solvable model for cubic symmetry. Nuovo Cimento D, 5:415–449
Calleja M, Nordstrom M, Alvarez M, Tamayo J, Lechuga LM, Boisen A, (2005) Highly sensitive
polymer-based cantilever-sensors for DNA detection. Ultramicroscopy, 105(1–4):215–222
Cocchini F, Bassani F, Bourg M () Model calculation of the optical properties of metallic particles
in a dielectric medium. Surf. Sci., 156:851–858
Coombes CJ (1972) Melting of small particles of lead and indium. J. Phys. F: Met. Phys.,
2:441–449
Dingreville R, Qu J, Cherkaoui M (2004) Effective elastic modulus of nano-particles, Proc. 9th
Int’l Symp. on Adv. Packaging Mat., pp. 187–192
Feynman RP (1959) Plenty of room at the bottom, APS meeting
Genzel L, Martin TP, Kreibig U (1975), Dielectric function and plasma resonances of small metal
particles, Zeitschrift fur Physik B, 21:339–346
Huang WC, Lue JT (1994) Quantum size effect on the optical properties of small metallic parti-
cles. Phys. Rev. B, 49 (24):279–285
Jin XC, Degertekin FL, Calmes S, Zhang XJ, Ladabaum I, Khuri-Yakub BT (1998) Micromachined
capacitive transducer arrays for medical ultrasoundimaging, Proc. Ultrasonics Symp.,
2:1877–1880
Kawabata A, Kubo R (1966) Electronic properties of fine metallic particles. II. Plasma resonance
absorption. J. Phys. Soc. Jpn, 21:1765–1772
Kuo DMT, Fang A, Chang YC (2001) Theoretical modeling of dark current and photoresponse
for quantum well and quantum dot infrared detectors. Infrared Phys. Technol., 42:433–442
Lai SL, Carlsson JRA, Allen LH (1998) Melting point depression of Al clusters generated during the
early stages of film growth: Nanocalorimetry measurements. Appl. Phys. Lett., 72:1098–1100
Lilleodden ET, Zimmerman JA, Foiles SM, Nix WD (2001) An experimental and computational
study of the elastic-plastic transition in thin films. Proc. Mat. Res. Soc. Symp., 673:1.3.1–1.3.6
Petersen KE (1982) Si as a mechanical material. Proc. IEEE, 70 (5):420–457
Pozhar LA (2000) Structure and dynamics of nanofluids: Theory and simulations to calculate
viscosity. Phys Rev E, 61 (2):1432–1446
Scaffardi LB, Tocho JO (2006) Size dependence of refractive index of gold nanoparticles.
Nanotechnology, 17:1309–1315
Scaffardi LB, Pellegri N, de Sanctis O, Tocho JO (2005) Sizing gold nanoparticles by optical
extinction spectroscopy. Nanotechnology, 16:158–163
Srikar VT, Spearing SM (2003) Materials selection in micromechanical design: An application of
the Ashby approach. J. MEMS, 12 1:3–10
Truong V, Courteau P (1987) Optical properties of very fine Al particles: Quantum size effect.
J. Appl. Phys., 62 (12):4863–4866
Wood DM, Ashcroft NW (1982) Quantum size effects in the optical properties of small metallic
particles. Phys. Rev. B, 25:6255–6274
Zhang M, Efremov MY, Schiettekatte F, Olson EA, Kwan AT, Lai SL, Wisleder T, Greene JE,
Allen LH (2000) Size-dependent melting point depression of nanostructures: nanocalorimetric
measurements. Phys. Rev. B, 62: 10548–10557
Zhang Z, Li JC, Jiang Q (2000) Modelling for size-dependent and dimension-dependent melting
of nanocrystals. J. Phys. D: Appl. Phys., 33:2653–2656
Chapter 2
Transduction Principles
Jeffrey Fortin
Abstract This chapter presents the most common fundamental transduction principles
used in microsensors. Each section provides an overview of the theory and then
gives an example of a sensor that uses the transduction principle being described. A
classification of measurands is presented as well as the most common transduction
techniques including piezoresistance, piezoelectricity, capacitive, resistive, tunneling,
thermoelectricity, optical and radiation-based techniques, and electrochemical.
Introduction
This chapter presents the most common fundamental transduction principles used
in microsensors. Each section provides an overview of the theory and then gives an
example of a sensor that uses the transduction principle being described.
Wikipedia defines a transducer as follows.
A transducer is a device, usually electrical, electronic, or electro-mechanical,
that converts one type of energy to another for the purpose of measurement or
information transfer. Most transducers are either sensors or actuators. In a broader
sense, a transducer is sometimes defined as any device that senses or converts a
signal from one form to another. (www.Wikipedia.com)
In a similar definition a transducer is defined as a device providing a usable output
in response to a specific measurand, where the measurand is defined to be the physi-
cal quantity, property, or condition that is to be measured (Norton 1982). It is further
stated here that when one is designing a sensor or trying to choose the appropriate
transduction technique there are a few questions one can ask, including: What is the
measurand? What is the principle of transduction? What is the sensing element?
What are the limits of the measurand to which the transducer will need to respond?
J. Fortin
GE Global Research Center. 1 Research Circle, KW C314, Niskayuna, NY12309
e-mail: fortinje@research.ge.com
18 J. Fortin
White has presented a classification scheme for measurands or properties one

may be interested in measuring. The main categories and significant subcategories
are given in Table 2.1 (White 1987).
Transducers are typically designed to sense a specific measurand and to ideally
respond only to that particular measurand. In reality a transducer will most likely
respond to the measurand in question and will also respond to other energy sources
that act on the sensor that are not of interest. These are considered sources of noise,
for example, when measuring strain with a piezoresistor the measurand of interest
is strain, however, the resistance will also change with temperature.
Table 2.2 shows the most common transducer types used to quantify the major
categories of measurand. The sections below provide an overview of the common
transduction mechanisms and principles such that one can begin to apply this
knowledge to sensor design.
Table 2.1 Classification of Measurands

Measurand Property of Interest
Acoustic Wave amplitude, phase, polarization Wave velocity Spectrum
Biological Identity, concentration, state
Chemical Identity, concentration, state
Electrical Current, charge potential, potential difference Field (amplitude,
phase, polarization) Conductivity and permittivity
Magnetic Field (amplitude, phase, polarization) Flux Permeability
Mechanical Position, velocity, acceleration Force Stress, strain Mass, den-
sity Flow Moment, torque Stiffness, compliance Viscosity
Crystallinity
Optical and Radiation Wave amplitude, phase, polarization Spectrum Velocity Energy
Thermal Temperature Flux Specific heat Thermal conductivity
Table 2.2 Transduction Techniques for Common Measurands

Most Common Transduction Techniques Utilized to Quantify
Measurand the Measurand
Acoustic Piezoelectric Piezoresistive Capacitive Optical
Biological Piezoelectric Piezoresistive Electrical Optical
Chemical Piezoelectric Piezoresistive Electrochemical Electrical Optical
Electrical Electrical Optical
Magnetic Piezoresistive Piezoelectric Electrical – capacitive, tunneling, Optical
Mechanical Piezoelectric Piezoresistive Capacitive Optical
Optical and Radiation Thermoelectric (Seebeck) Photosensitivity (photovoltaic, photoelec-
tric, photoconductors, photodiodes, and phototransistors)
Thermal Thermoelectric Photosensitivity Electric – resistive
2 Transduction Principles 19
Piezoresistivity
Piezoresistivity is, in its most basic form, the change in a material’s resistance
resulting from a change in stress in the material. The word piezo is derived from
the Greek word piezein, which means to press or squeeze. Many materials exhibit
the piezoresistive effect and it is typically quantified by what is termed the gauge
factor. The gauge factor is the change in resistance per given strain per starting
resistance and can be described via the following equation,
DR
GF = . (2.1)
R.Strain
The gauge factor for silicon decreases with increasing impurity concentrations and
this can be predicted by model. Controlled doping is typically accomplished via ion
implantation of the specific doping ion into the silicon to define the piezoresistor.
An alternative technique is to deposit a film containing the doping ion over a pat-
terned Si surface and then use a temperature treatment to drive the dopant into the
depth of the Si.
The piezoresistive gauge factor decreases as temperature increases, and this can
also be predicted. The coefficients increase linearly with the inverse of temperature.
One can find a deep description of the mathematics in Sze (1994).
When deciding if piezoresistance can be used as a transduction measurand for a
particular measurand one only has to determine if the sensor can be designed such
that the measurand can produce a stress on a portion of the device where a piezore-
sistor can be located. Examples of measurands that are quantified via piezoresist-
ance include: pressure, vibration, acceleration, and magnetic field. Once it has been
decided that the measurand could be quantified via piezoresistance one must deter-
mine if it is the best approach that will meet all the specifications of the application,
as described in Chapter 1.
A great example of a transducer that uses the piezoresistive effect is a silicon-
based MEMS pressure sensor. Silicon-based pressure sensors have been around
since the late 1950s and they are a very mature technology. GE Sensing offers a
multitude of Si-based pressure sensors for a variety of applications including blood
pressure sensing, tire pressure sensing, industrial process measurement, and so on.
An overview of one pressure sensor is presented here that is designed to measure
tire pressure (www.GESensing.com).
This sensor is approximately 1 by 1 mm and is an absolute pressure sensor,
meaning the reference cavity is a vacuum and, once calibrated, the sensor gives the
absolute pressure inside the tire. As with most pressure sensors, the electronic readout
technique utilizes a Wheatstone bridge.
The sensor element is a thin silicon membrane with embedded piezoresistors.
The piezoresistors are formed via an implantation step and the proper doping level
is chosen to provide the highest gauge factor. The piezoresistors are positioned in
the areas of the membrane that see the highest strain due to the force of the pressure
bending the membrane (Fig. 2.1).
20 J. Fortin
Diaphragm Wire -Bond Pad
Reference Cavity
Piezo- Resistor
Fig. 2.1 Side view of GE Sensing silicon pressure sensor. (Courtesy of GE Sensing.)
P1 Applied
Pressure
Reference Cavity
P2
Pressure
Fig. 2.2 Side view representing diaphragm deflection of GE Sensing silicon pressure sensor.
(Courtesy of GE Sensing.)
Many membranes are formed on a silicon wafer using typical microfabrication

processes and then this wafer is wafer-bonded in a vacuum environment to a bottom
silicon wafer with cavities lined up with the membrane. This forms a drumlike
structure with a vacuum cavity lying below a thin membrane. As the external pres-
sure fluctuates, the membrane moves to balance the force between the external
pressure and the stretched membrane (Fig. 2.2).
Piezoelectricity
Overview of Theory. Piezoelectricity, as is piezoresistivity, is an electrical effect

caused by a change in the strain of a material. In the cause of piezoelectricity, when
a piezoelectric material is stressed (compressive or tensile) a charge is induced
across the material’s faces in response to the magnitude and direction of the strain.
+ + + + + + + + + + + + + + + +
− − − − − − − − − − − − − − − −
(a) (b)
Fig. 2.3 (a) Piezoelectric material with no externally applied strain has no net charge on its
surfaces. (b) The same material under strain produces charges at surface that can be measured
A piezoelectric transducer therefore converts a change in a measurand into a change

in electrostatic charge or voltage (Fig. 2.3).
Typical piezoelectric crystal materials include LiNbO3, LiTaO3, Li2B4O7, GaAs,
and quartz. Typical thin film-based piezoelectric materialso include ZnO, AlN, and
PZT (Pb(Zr, Ti)O3). PZT is probably the most widely used piezoelectric material
for sensing in applications such as accelerometers, vibrometers, ultrasound, and
high dynamic range ac pressure sensing.
In microsensors, piezoelectric materials can be either directly deposited onto the
device or they may be integrated into the device, for example, lamination of a piezo-
electric polymer film. Because PZT has an order of magnitude higher piezoelectric
effect than ZnO and AlN many techniques have been developed to integrate PZT with
a microdevice, including sol–gel and sputtering. ZnO and AlN films have also been
integrated into microdevices for the purpose of transduction (Royer et al. 1983; Ried
et al. 1993; Ko et al. 2003). The definition of methods and measurement of piezoelec-
tric crystal units is reported in detail in Halfner (1969).
An example microdevice utilizing a piezoelectric transduction technique was
recently presented at the 2006 Transducers Conference at Hilton Head (Horowitz
et al. 2006). The device presented was a micromachined piezoelectric microphone
for aeroacoustics applications. Although there had been previous research done on
MEMS-based microphones most of them have been developed for audio applica-
tions. The microphone reported at 2006 Hilton Head was designed for high sound
pressure applications in excess of 160 dB with a bandwidth of >50 kHz.
The microphone was fabricated by combining a sol–gel PZT (lead zirconate–
titanate) deposition process on a silicon-on-insulator wafer. The PZT was deposited
onto a 1.80 mm diameter 3 mm thick Si diaphragm. The PZT was processed and
lithographically defined to an annular ring near the diaphragm edge to maximize
the sensitivity. The PZT layer was 270 nm thick and was placed between two thin
metal electrodes. A diffusion barrier separated the PZT from the silicon diaphragm.
As acoustic energy impinged upon the diaphragm it moved. As it moved, the piezo-
electric material experienced stress in the z-axis and the charge at its surfaces
changed in response to this stress and this charge was measured.
22 J. Fortin
Test results for the microphone showed a sensitivity of 0.75 uV/Pa with a linear
dynamic range from 47.8 to 169 dB and a resonant frequency of 50.8 kHz.
Electrical—Resistance, Capacitance, Impedance, Tunneling
There are many ways to convert a change in an external measurand into a change
in an electrical signal directly, particularly in a MEMS device, where there are moving
parts that can be accessed electrically or integrated thin films that often have electrical
properties based on the environment to which they are exposed.
Resistance: One key transduction technique that converts the measurand to a
change in resistance is piezoresistance, which, because of its widespread use in
microdevices was covered separately above. Other resistance-based transduction
techniques also rely on the measurand interacting with a film or bulk structure and
hence changing an electrical property. For example, certain polymers have a moisture-
or gas-sensitive resistivity. Another example is in the measurement of temperature,
where because resistance of a material is a function of temperature it can be directly
used to measure temperature; this type of device is referred to as a thermistor.
An example of a device that uses a materials change in resistance due to expo-
sure to the measurand can be seen in the work of Valentini et al. (2004). This device
uses carbon nanotubes (CNTs) as the functional transducer material. Carbon nano-
tubes present extremely high surface-to-volume ratios and have recently seen
significant attention for their gas adsorption properties (Treacy 1996). Valentini et al.
used an interdigital electrode structure made from platinum deposited and patterned on
top of a silicon nitride film. The CNTs were then grown from a catalyst between
the Pt electrodes to heights of approximately 200 nm. The results showed that the
resistance of the film on CNTs decreased when in contact with NO2 and increased
when in contact with NH3, ethanol, water vapor, and C6H6. The detection limit for
NO2 was shown to be as low at 10 ppb.
Capacitance: In this transduction technique the measurand interacts with the
device to change the capacitance value of a capacitor. This change can be induced
by changing the effective distance between the two plates or electrodes of the
capacitor or by changing the dielectric constant of the insulator material. Examples
of both are given below.
The capacitive transduction technique that can be used to measure pressure is
capacitance. In a typical silicon-based capacitive pressure sensor the design is similar
to a piezoresistive pressure sensor as described above. Instead of implanting pie-
zoresistors into the diaphragm, the diaphragm itself is used as one plate of a capaci-
tor. Alternatively a metal layer can be placed on or embedded in the diaphragm. The
second plate or electrode is located at the bottom of the gap (see Fig. 2.5). In the
sensor shown in Fig. 2.4 the substrate is a degenerately doped silicon wafer and the
membrane has been wafer-bonded under vacuum to a patterned oxide layer. Contact
is made to the substrate through an opening in the oxide and directly to the also
Pressure
Reference
Ground Capacitor
Si diaphragm
Si Wafer
Fig. 2.4 Capacitive-based pressure sensor
G-CAP2, calibrated, 10 kHz, 1 VAC, 25⬚C (77⬚F)

190
Capacitance, pF
180
170
160
150
140
0 10 20 30 40 50 60 70 80 90 100
%RH
(a) (b)
Fig. 2.5 (a) G-CAP2TM Humidity Sensor; (b) response curve for G-CAP2TM capacitive humidity
sensor. (Courtesy of GE Sensing.)
degenerately doped silicon membrane. A capacitor with an oxide dielectric as a

reference is placed next to the sensor.
As the external pressure changes the diaphragm moves and the distance between
the diaphragm and the lower electrode changes, thus changing the capacitance of
the device. One advantage of a capacitive-based pressure sensor over a piezoresis-
tive approach is the lower power consumption of the sensor itself. Another advan-
tage is the capacitive approach tends to have higher sensitivity if properly designed
(Eaton and Smith 1997).
The G-Cap Moisture Sensor offered by GE Sensing is an example of a device
that utilizes the change in capacitance of a thin film in response to the measur-
and. In this case the functional material is thin polymeric film sandwiched
between two thin patterned electrodes. The polymeric film was developed to
allow for measurement of a wide range of relative humidity from 0 to 100% and
it can survive total immersion in water without loss of accuracy. The typical
capacitance of the sensor is in the range of 140–190 pF and it changes linearly
with %RH. The capacitance is measured between 1 kHz and 1 MHz (Fig. 2.5).
The sensitivity to changes of temperature can be calibrated and is typically less
than 0.05% RH/°F.
24 J. Fortin
Impedance: This is very similar in nature to the resistance or capacitance tech-

nique. In impedance-based transduction the AC impedance of a component of the
device is measured and monitored. Often there is a material that is used to measure
an environmental parameter, such as humidity or local magnetic field.
One impedance-based transduction method that has received a lot of atten-
tion in recent years is giant magneto impedance (GMI). This effect refers to a
material’s very large change in resistance due to an applied magnetic field. This
effect has been seen in amorphous wire, ribbons, and thin films. An analogous
effect is called giant stress impedance (GSI), where a change in stress causes a
large change in impedance. The GMI effect is seen in many high-permeability
materials such as amorphous soft magnetic wires with Fe-based or Co-based
compositions (Zribi et al. 2005; Han et al. 2005; Garcia et al. 2005). This effect
depends on the material’s permeability which is a function of many factors
including domain configurations, material geometry, anisotropy, and excitation
frequency. The GMI effect can be expressed as an impedance ratio given by
Equation (2.2),
DZ / Z = [ Z ( H ) − Z ( H max)] / Z ( H max), (2.2)
where Z is the impedance, ΔZ is the change in impedance, H is the magnetic field

applied to the sensor, and Hmax is the max field. GSI can be expressed in a similar
fashion where the magnetic field is replaced with stress.
Han et al. (2005) provide a nice overview of sensors utilizing the GMI and GSI
effect. The applications of these sensors include magnetic field measurement, a
position measurement for the location of a catheter in the human body, nondestruc-
tive testing, and electronic surveillance. Zribi et al. (2005) also present an oil-free
stress impedance pressure sensor for harsh environments.
Tunneling: Another interesting technique employed in micro- and nanosys-
tems for transduction of a measurand value into an electrical signal is electrical
tunneling. Typically what is done in this case is a mechanical component
becomes one electrode in a two-electrode circuit. This mechanical component
will have a tip or surface placed within a few to tens of nanometers from the
second stationary electrode. The mechanical component with the tip will move
in response to the measurand. As it moves, the distance between the tip and the
stationary electrode will change and hence the tunneling current will change.
This technique can be used to measure very tiny changes in the measurand if
properly designed.
A number of publications have addressed the design, fabrication, and perform-
ance of tunneling-based sensors. Liu and Kenny (2001) have demonstrated a
MEMS-based high-precision, wide-bandwidth micromachined tunneling acceler-
ometer with a resolution of 20 ng/sqrt Hz and 5 Hz–1.5 kHz bandwidth. The design
consists of a cantilever tip substrate, a proof mass, substrate, and a cap substrate all
wafer-bonded together. The accelerometer is operated at a pressure of 10 mTorr to
reduce thermomechanical noise and increase Q to above 100. A feedback controller
is used to maintain the tunneling gap at 10 Å.
Thermoelectricity
Overview of Theory
Thermoelectricity: This transduction technique converts the value of the measurand

into a voltage (or electromotive force) generated by the potential difference between
the junctions of two selected dissimilar materials due to the Seebeck effect. The
Seebeck effect is well known and is the technology employed in thermocouples,
where, as the temperature of the junction changes, the voltage across the junction
changes. Thus, a thermocouple works by measuring the difference in potential
caused by the dissimilar wires. Several thermocouples in series are called a thermo-
pile. This technique is also used in silicon-based devices to measure temperature
using a noncontact approach. The Seebeck effect can be described by Equation
(2.3) referring to Fig. 2.6:
V = (SB − S A ) / (T2 − T1 ), (2.3)

where SA and SB are the Seebeck coefficients of materials A and B, respectively and
T1 and T2 are the temperatures of the two junctions.
A good example of a microsensor utilizing the Seebeck effect for transduction
is the GE silicon-based IR thermopile (www.GESensing.com). The IR thermopile
sensor consists of a number (about 40) of thermocouple pairs connected in series
and covered with a high emissivity coating. The hot junctions are thermally isolated
from the cold junctions and are exposed to the incident IR radiation. The cold junc-
tions are attached to a heat sink. GE’s device is built on a silicon wafer and the
thermocouple radiation detection junctions are placed on a thin, low thermal mass
diaphragm, and the reference junctions are placed off the membrane on the thick
silicon wafer (Figs. 2.7 and 2.8). A thermistor is placed in the finished package as
a reference.
The IR thermopile device allows for measurement of temperature without direct
contact with fast, millisecond response times due to the low thermal mass of the
diaphragm. Applications include tympanic thermometers for body temperature
measurement, food temperature measurement in microwave ovens, measuring
temperatures inside vehicles, and many others.
Wire A
Wire B
T2 T1
V
Fig. 2.6 Thermocouple consisting of wire of material A connected to wire of material B

26 J. Fortin
REFERENCE
JUNCTIONS
RADIATION
DETECTING
JUNCTIONS
RADIATION
ABSORBER
Fig. 2.7 Arrangement of thermocouple junctions on the silicon IR sensor
Thermoelement 1 Black Body

Thermoelement 2
Cold Junction
Diaphragm Hot Junction
Fig. 2.8 Cross-section of the GE thermopile IR sensing chip. (Courtesy of GE Sensing, www.
GESensing.com.)
Optical and Radiation Techniques
Overview of Theory
Optical techniques can be used to measure the change in a mechanical measur-

and, or quantify an optical spectrum for chemical or biological analysis using IR,
UV, or other radiation. The optical sensing techniques typically probe the mate-
rial/object to be sensed directly or indirectly via some structure. In all optical
techniques a detector must quantify the optical signal. Typically some property of
the optical beam returning from the sample is compared to the source beam,
including amplitude and phase information. A radiation-sensitive detector can be

used either combined with a radiative probe (e.g., Raman spectroscopy) or to meas-
ure directly a radiative property of the sample/source being measured (e.g., gamma
detection).
Electromagnetic or radiant energy is carried by photons spanning wavelengths
of cosmic rays from 10−9 um to radio waves of 1014 um long. Electromagnetic radia-
tion primarily interacts with semiconductors via absorption processes and interfer-
ence, diffraction, reflection, polarization, transmission, and refraction all play a role
(Sze 1994). In the measurement of radiation there are many different transduction
techniques that can be used including photovoltaic, photoelectric, photoconductors,
photodiodes, and phototransistors.
In a photovoltaic or photodiode light is incident upon a junction between
dissimilar materials and a voltage is generated. This technique can be used to meas-
ure the intensity of a light source and is also used in power generation. In photo-
conduction a change in measurand is converted to a change in the resistance or
conductance of a semiconductor material due to a change in the amount of illumi-
nation incident upon the material. The resistance or conductance can then be meas-
ured electronically. In the photoelectric effect an incident photon causes the
emission of an electron. A common detector for optical imaging or sensing is the
charge-coupled device (CCD).
There is a variety of different sensors that utilize optical or radiation techniques
to quantify a measurand. One example is a fiber optic strain sensor, which utilizes
a fiber optic with embedded Bragg gratings. The distance between the Bragg grat-
ings changes as the fiber is put under stress or strain and this distance can be meas-
ured as an optical beam. This fiber technique is available for temperature and strain
and one fiber can have multiple sense locations allowing for multiplexing multiple
measurements.
One interesting example of optical sensing where functional nanostructures play
a role is surface-enhanced Raman spectroscopy (SERS). In this technique the
Raman signal generated from a sample can be greatly enhanced (106–108) if the
sample is in close proximity (nm range) to an appropriately roughened surface
(Moskovits 2005). This technique has been used for biosensing applications via
using nanoparticles as the SERS substrate and selectively attaching pathogens to
the nanoparticles using an antibody approach (Stuart et al. 2005).
Electrochemical
Overview of Theory
There are many variations of approaches to utilizing electrochemistry to make meas-

urements and transducer signals using microsensor technology. The classifications
of electrochemical microsensors include: potentiometeric (measurement of
28 J. Fortin
potential) and amperometeric (measurement of current). The devices utilized are

typically one of the following: chemoresistor, chemocapacitor, chemodiode, chem-
otransistor, or CHEMFET (Gardner 1994).
This transduction technique can be used to measure or quantify materials in the
gas or liquid phase, and includes chemical and biological measurands such as
sodium ion concentration, oxygen concentration, and glucose measurement. The under-
lying principle is to utilize an electrochemical reaction between the species of inter-
est and the functional material on the device and to measure the degree of that
interaction via the measurement of an electrical property of the active layer in the
device or the direct electrical signal that is produced from the reaction.
Gardner gives a good overview of the measurement of chemical and biological
sensing using the electrochemical approach (Gardner 1994).
The potentiometric technique is described here to give a better understanding of
electrochemical transduction. In this technique the potential difference between an
indicator or reference electrode and the sample electrode is measured. This approach
is prevalent in macroscopic systems today and scales well to microsensors because
the magnitude of the potential does not change with the size of the electrode. The
typical implementation of a potentiometric microsensor is via a CHEMFET. In a
CHEMFET the gate of a field effect transistor (FET) can be chemically modulated
and gives rise to three types of microsensors: the ion-sensitive FET (ISFET), the
work function FET, and the enzymatically selective FET (ENFET) (Janata 2003).
ISFETS are well adapted to the measurement of ions in aqueous solutions and
often utilize polymer-based ion sensitive layers such as polyvinlychloride, poly-
HEMS/siloprene, polyurethane/acrylate, and polysiloxane (Humenyuk et al. 2006).
Humenyuk et al. recently published on the development of pNH4-ISFETS for water
analysis using a polysiloxane-based ion selective layer. The device was fabricated
on N-type silicon using standard P-well technology. The gate structure was a
LPCVD deposited 80 nm Si3N4 layer with a polysiloxane copolymer deposited on
top. A reference metal oxide/nitride FET was fabricated on each die for drift and
temperature compensation. Sensitivity was shown to be around 47 mV/pNH4
through the concentration range of 1–5 pNH4.
Summary
Fundamentally transduction is taking energy from one form and transferring it into
another and quantifying that energy change or energy input. As one can see there are
a number of measurands that can be quantified via a variety of transduction
techniques. There are often multiple transduction approaches to quantify a measur-
and and one must fully understand the specifications of the application in order to
down-select approaches. The final approach may not necessarily be clear and experi-
mentation or innovation may be required to determine or define the best approach.
References
Eaton WP, Smith JH (1997) Micromachined pressure sensors: review and recent developments.
Smart Mater. Struct., 6:530–539
Garcia C, Zhukov A, Zhukova V, Ipatov M, Blanco JM, Gonzalez J (2005) Effect of tensile stresses
on GMI of Co-rich amorphous microwires. IEEE Trans. Magnetics, 41(10):3688–3690
Gardner JW (1994) Microsensors: Principles and Applications. Wiley, West Sussex, England
Han M, Liang DF, Deng LJ (2005) Review paper, sensors development using its unusual proper-
ties of Fe/Co-based amorphous soft magnetic wire. J. Mat. Sci., 40:5573–5580
Halfner E (1969) The piezoelectric crystal unit-definitions and methods of measurement. Proc.
IEEE 57, No. 2
Horowitz S, Nishida T, Cattafesta L, Sheplak M (2006) Solid State Sensors, Actuators, and
Microsystems Workshop, Hilton Head Island, SC, June 4, 2006
Humenyuk I, Torbiero B, Assie-Souleille S, Colin R, Dollat X, Franc B, Martinez A, Temple-
Boyer P (2006) Development of pNH4-ISFETS microsensors for water analysis.
Microelectronics J., 37:475–479
Janata J (2003) Electrochemical microsensors. Proc. IEEE, 91(6):864–869
Ko S, Kim Y, Lee S, Choi S, Kim S (2003) Micromachined piezoelectric membrane acouostic
device. Sensors and Actuators A, Physical, 103:130–134
Liu CH, Kenny TW (2001) A high-precision, wide-bandwidth micromachined tunneling acceler-
ometer. J. MEMS, 10(3):425–433
Moskovits M (2005) Surface-enhanced Raman spectroscopy – A brief retrospective. J. Raman
Spectroscopy, 36(6/7):485–496
Norton H (1982) Sensor and Analyzer Handbook. Prentice Hall, NJ, pp. 18–24
Ried R, Kim E, Hong D, Muller R (1993) piezoelectric microphone with on-chip CMOS circuits.
J. MEMS, 993 (23):111–120
Royer M, Holmen J, Wurm M, Aadland O (1983) ZnO on Si integrated acoustic sensor. Sensors
and Actuators, A: Physical, 4:357–362
Stuart DA, Haes AJ, Yonzon CR, Hicks EM, Van Duyne RP (2005) Biological applications of
localised surface plasmonic phenomenae. IEE Proc. – Nanobiotechnol., 152(1):13–32
Sze SM (1994) Semiconductor Sensors. Wiley, New York
Exceptionally high Young’s modulus observed for individual carbon nanotubes”, Treacy, M.M.J.
(NEC Res. Inst., Princeton, NJ, USA); Ebbesen, T.W.; Gibson, J.M., Nature, v 381, n 6584, 20
June 1996, p 678–680
Valentini L, Cantalini C, Armentano I, Kenny JM, Lozzi L, Santucci S (2004) Highly sensitive
and selective sensors based on carbon nanotubes thin films for molecular detection. Diamond
Relat. Mat., 13:1301–1305
White RM (1987) A sensor classification scheme. IEEE Trans. Ultrason. Ferroelec, Freq. Contr.
UFFC-34:124
Zribi A, Iorio L, Lewis D (2005) Oil-free stress impedance pressure sensor for harsh environ-
ments. IEEE. Vol 2005, p 1275–1277
Chapter 3
Growth and Synthesis of Nanostructured
Thin Films
Yiping Zhao
Abstract Nanostructured thin film (NSTF) is composed of thin layers of nano-

structured objects such as nanoparticles, nanorods, nanotubes, nanowires, and nano-
porous networks. Fabrication and synthesis of those nanostructured thin films are
essential for exploring their properties and creating advanced applications. This
chapter gives an overview of a range of synthesis methods for NSTFs, such as
thermal vapor transport methods, catalyst-assisted fabrication methods, physical
vapor deposition methods, chemical vapor deposition methods, sol–gel methods,
diblock copolymer methods, spin coating methods, electrochemical deposition/etching
methods, electrospinning methods, and template-based synthesis techniques. In
the end, we have detailed an emerging nanofabrication method, the glancing angle
deposition method, and its capability to design NSTF with different geometry and
compositions.
Introduction
Fabrication and synthesis of nanostructured thin films (NSTFs) are essential for
exploring their properties and creating advanced applications. In general, the nano-
structured thin film (NSTF) is defined as an assembly of a thin layer of nanostruc-
tured objects such as nanoparticles, nanorods, nano-tubes, nanowires, nanoporous
networks. It is different from individual or bundles of nanometer scale objects. In the
literature, there are two kinds of nanostructured thin films (NSTFs) that have been
used without any differentiation. One refers to an ultrathin film with submicrometer
thickness (£100 nm), either having a continued morphology or a discontinued
islandlike morphology. This kind of NSTF is usually prepared by conventional
thin-film deposition techniques such as physical vapor deposition (PVD), chemical
Y. Zhao
University of Georgia, Department of Physics and Astronomy, 221 Riverbend Research
South Laboratory, 220 Riverbend Road, Athens, GA 306028
e-mail: zhaoy@physast.uga.edu
32 Y. Zhao
Table 3.1 Process Conditions and NSTF Structures for NSTF Fabrication or Synthesis
Fabrication or Synthesis Conditions NSTF Structures
Approaches: bottom-up or top-down Morphology: nanoparticle, nanorod, nanotubes,
nanowires, nanoporous network, or mixtures
Mechanism: physical or chemical process Orientation: aligned or misaligned
Environment: gas or liquid phase Regularity: ordered or disordered
Temperature: high or low Layered structure: single layer or multilayer,
single material or different materials
Other factors, such as pressure, carrier gas,
plasma, substrate, etc.
vapor deposition (CVD), or electrochemical deposition (ECD). The other NSTF

refers to a film consisting of nanostructured objects such as nanoparticles, nanorods,
nanotubes, nanowires, nanoporous network, and so on. Compared to the first kind
of NSTF, the morphology is more complicated, and the fabrication or synthesis
techniques are very diverse.
The conventional thin-film deposition techniques mentioned above still play
important roles, although there have been a number of other nanofabrication or
nanosynthesis techniques developed within the past 20 years that are becoming
more and more significant. In fact, the fabrication of NSTFs is compatible with
nanofabrication techniques in general, and we can catalogue the fabrication tech-
niques according to top-down and bottom-up approaches. The top-down approach
refers to machine or carved macroscopic structures down to nanometer scales,
whereas the bottom-up approach refers to controlled or directed self-assembly of
atoms and molecules into nanostructures. The top-down approach usually consists
of at least one lithography step or another size definition step and one etching step
that is discussed in detail in the next chapter. Here we concentrate on how to use
the bottom-up approach to fabricate NSTFs. Because there is a host of nanofabrication
techniques and a variety of NSTF morphologies for each individual fabrication
technique, we need to pay attention to a number of parameters listed in Table 3.1
in order to further integrate the process with other nano- or microfabrication proc-
esses for device fabrication, as discussed in Chapter 4.
We organize the fabrication methods according to their growth conditions. In the
first three sections, we discuss different NSTF fabrication techniques in the gas
phase and the liquid phase and through templates. In the last section, we give a
detailed description of a gas phase nanofabrication technique called the glancing
angle deposition (GLAD) method.
Gas Phase Fabrication Techniques
Gas phase fabrication methods usually represent a clean environment fabrication or

synthesis because most NSTF fabrications are under vacuum. According to the
general synthesis conditions, we discuss four major synthesis methods: the thermal
3 Growth and Synthesis of Nanostructured Thin Films 33
vapor transport method, the catalyst-assisted fabrication method, the physical vapor
deposition method, and the chemical vapor deposition method.
Thermal Vapor Transport Methods
The thermal vapor transport method is used for depositing nanorod or nanowire thin
films directly from the vapor of the materials under the supersaturating condition
(Xia et al. 2003; Wang 2004). The deposition can be a physical process through
nucleation, and it can also involve chemical reactions during deposition such as
oxidation. The experimental setup for a typical thermal vapor transport deposition
system is shown in Fig. 3.1.
The system consists of at least a furnace to control the temperature of the fabri-
cation chamber, a source to provide the material, and a substrate to collect the
NSTF. In most cases, a carrier gas is introduced to transport the vapor from the
source to the substrate. Also a vacuum pump is used so that a desired chamber pres-
sure can be obtained. In some cases, neither the carrier gas nor the vacuum pump
is used. The source vapor can be introduced directly from the crucible into the
chamber or by other thermal sources outside the chamber (by carrier gas) or by
other techniques such as laser ablation, chemical reduction, and so on.
The important processing parameters are temperature (both the chamber and the
substrate), the partial pressures of the carrier gas and the vapor source, the flow and
type of carrier gas, and the growth time. Usually the substrate temperature is lower
than the source temperature. Fig. 3.2 shows several examples of nanorod thin films
fabricated by the thermal vapor transport methods (Huang et al., 2001; Yang et
al. 2002). Under most growth conditions, randomly aligned and networklike nano-
rod or nanowire thin films are formed. With fine-tuned growth conditions, well-
aligned nanorod arrays can be fabricated. In general, the growth temperature is
relatively high.
Furnace
Source materials Growth substrate
Carrier Gas To Pump

(Ar or N2)
Fig. 3.1 A general experimental setup for thermal vapor transport method (Wang 2004).
(Reprinted by permission from the Annual Review of Physical Chemistry, Volume 55, copyright
2004, Annual Reviews, www.annualreviews.org)
34 Y. Zhao
Fig. 3.2 (a) Randomly ordered ZnO nanorods array (Huang et al. 2002b). (Reprinted by permis-
sion from Advanced Materials 13, 113 (2001), copyright 2001, Wiley.) (b) Aligned ZnO nanorod
array (Yang et al. 2002) (Reprinted by permission from Advanced Functional Materials, 12, 323
(2002), copyright 2002, Wiley) fabricated through thermal vapor transport method
(a) (b)
300 nm
Fig. 3.3 Aligned ZnO nanorod array fabricated through catalyst assisted method (a) Au catalyst
(Ng et al. 2003) (Reused with permission from Hou Tee Ng, Applied Physics Letters, 82, 2023
(2003) copyright 2003, American Institute of Physics), and (b) nio catalyst (Lyu et al. 2002).
(Reprinted by permission from Chemical Physics Letters 363, Seung Chul Lyu, Low temperature
growth and photoluminescence of well-aligned zinc oxide nanowires, 134, copyright 2002,
Elsevier)
Catalyst-Assisted Fabrication Methods
The catalyst-assisted nanorod fabrication method is a variation of the thermal vapor

transport method requiring the introduction of a catalyst layer onto the substrate or
the injection of a small amount of precursor that can be decomposed into catalyst
particles. The experimental setup is similar to Fig. 3.1. Fig. 3.3 shows two examples
of the ZnO nanorod films grown by this technique (Ng et al. 2003; Lyu et al. 2002).
Fig. 3.3a shows the ZnO nanorod array grown on an Au catalyst sapphire substrate
(Ng et al. 2003), and Fig. 3.3b displays the ZnO array on a NiO catalyst Si substrate
(Lyu et al. 2002).
One well-known catalyst-assisted fabrication method is the vapor–liquid–solid
(VLS) method (Lieber 1998). This method has been applied for whisker growth
Fig. 3.4 (a) Schematic illustration of the VLS process; (b) vertical Si nanowire array grown on
a (111) Si wafer (Wu et al. 2002). (Reprinted by permission from Chemistry-A European Journal,
8, 1260 (2002), copyright 2002, Wiley)
Fig. 3.5 A schematics illustrating the general steps and physical mechanism for a PVD process
since the 1960s (Levitt, 1970), and it employs a catalyst to promote the anisotropic
crystal growth. The catalyst usually forms eutectic droplets with the reactants, and
the size of the eutectic droplet determines the size of the nanorods. The eutectic
alloy droplet will sit on top of the nanorods until the rods cease growth due to other
reasons such as poisoning or end of growth as shown in Fig. 3.4a (Wu et al. 2002).
Fig. 3.4b also shows an example of an aligned Si nanorod array on a Si (100) sur-
face fabricated by the VLS method (Wu et al. 2002). In general, a specific catalyst
must be chosen for each material, and usually the growth temperature is relatively
high.
Physical Vapor Deposition Methods
Physical vapor deposition (PVD) is a conventional process for thin-film deposition.

As shown in Fig. 3.5, the PVD process involves at least the following steps: (1) the
material to be deposited is converted from condensed phase into vapor phase by
36 Y. Zhao
Fig. 3.6 Three major processes for PVD: (a) thermal evaporation, (b) sputtering growth, and
(c) pulsed laser deposition
physical means, (2) the vapor transports from its source to the substrate, and (3) the
vapor condenses on the substrate to form the thin film. There are several ways to
resolve the key step for the PVD process, that is, to convert the solid phase into the
gas phase. In general, three major processes have been utilized: (1) the thermal
evaporation method, (2) the sputtering growth method, and (3) pulsed laser deposition.
The basic experimental setups for these three fabrication techniques are illustrated
in Fig. 3.6.
For the thermal evaporation method, the source material is placed into a crucible,
and the crucible can be heated either by resistance or by an electron beam to its
melting temperature so that there is enough vapor coming from the source that can
be deposited onto the substrate. For the sputtering growth, a source target is used,
and inside the vacuum chamber high-energy ions are generated by ionization.
Those high-energy particles collide with the source target and knock out the atoms
or molecules from the source target. The ejected atoms or molecules condense onto
a substrate to form a thin film or NSTF. A similar idea is applied for pulsed laser
deposition; instead of using high-energy ions, a high-intensity, short-wavelength,
pulsed laser beam is used to bombard the target.
In most cases, the PVD methods are used to produce islandlike ultrathin films
especially with films of Au, Pt, or Ag for plasmonic applications (Bartlett et al.
2004), or to produce nanocluster films by co-deposition of metal and dielectrics
(Biswas et al. 2003, 2004). Recently, a so-called glancing angle deposition (GLAD)
technique has been developed to fabricate aligned nanorod array structures, and the
details of GLAD are discussed in the section, “Glancing Angle Deposition.”
Chemical Vapor Deposition Methods
The chemical vapor deposition (CVD) technique is a widely used method for
depositing thin films and nanostructures for a large variety of materials. In a typical
CVD process, as shown in Fig. 3.7, reactant gases or precursors diluted in a carrier
gas at room temperature enter the reaction chamber. The gas mixture is heated as it
Fig. 3.7 A schematic of a conventional CVD reactor
Fig. 3.8 Aligned multiwall carbon nanotube arrays fabricated through CVD technique (Wei et al.
2002). (Reprinted by permission from Natures 416, B. Q.Wei, R. Vajtai, Y. Jung, J. Ward, Y.
Zhang, P.M. Ajayan, G. Ramanath, Organized Assembly of Carbon Nanotubes, copyright 2002,
Macmillan)
38 Y. Zhao
Fig. 3.9 The two basic steps for ALD
approaches the deposition surface. Depending on the process and operating conditions,
the reactant gases may undergo homogeneous chemical reactions in the vapor phase
before striking the surface, and the reactions result in a condensed phase thin film form-
ing onto the substrates with the corresponding volatile reactants being pumped from the
chamber. In order to promote the chemical reaction, either a plasma or laser beam can
be introduced during the process. CVD primarily is a thin-film deposition technique.
With the help of different catalysts, the CVD has become a major technique to make
NSTFs over the past 20 years. For example, many different nanostructures such as
carbon nanotube arrays (Fig. 3.8; Wei et al. 2002), inorganic oxide nanorods or
nanowires, and others can be fabricated using the CVD technique.
A unique CVD process is called the atomic layer deposition (ALD) process. ALD
is based on sequential, self-limiting surface reactions as illustrated in Fig. 3.9 (Kim
2003). For one growth cycle, only one layer of precursors adsorbs onto the substrate
due to self-limiting surface adsorption, and then a reduction step is performed leaving
a monolayer of the desired atom/molecule on the substrate. Thus, one can deposit one
layer of atoms or molecules onto the substrate. Because the process is controlled by
surface adsorption, this unique growth technique can provide atomic layer control and
allow conformal films to be deposited on very high aspect ratio structures.
Liquid Phase Fabrication Techniques
Liquid phase fabrication techniques refer to the synthesized NSTFs under a wet envi-
ronment. These processes usually occur at relatively low temperatures but involve
chemical reactions. In the following section, several liquid phase fabrication methods
are discussed, but due to the limited knowledge of the author, there may be other
methods that have not been included.
Sol–Gel Methods
The sol–gel process is a versatile solution process for making ceramic and glass
NSTFs (Caruso and Antonietti 2001; Pomogailo 2005). The sol–gel process
involves the transition of a system from a liquid “sol” into a solid “gel” phase. The
“sol” is made of solid particles with a diameter of few hundred nanometers that are
suspended in the liquid phase. The particles condense into a new phase (gel) in
which a solid macromolecule is immersed in a liquid phase (solvent). A common
example is the hydrolysis and condensation reactions of metal alkoxides to form
larger metal oxide molecules. After a series of hydrolysis and polymerization reac-
tions, the precursor forms a colloidal suspension, or a “sol”. Further processing of
the “sol” enables one to make ceramic materials in different forms. Thin films can
be produced by spin-coating or dip-coating, and nanoporous NSTFs can be formed
after removing the wet “gel”. With proper viscosity of a “sol”, ceramic fibers can
be drawn from the “sol” through the electrospinning process discussed in the sec-
tion, “Electrospinning Methods.”
Diblock Copolymer Methods
The diblock copolymer method is very similar to the sol–gel method. A diblock
copolymer molecule consists of two immiscible polymer blocks A and B that
are covalently bonded at one end as shown in Fig. 3.10 (Harrison et al. 1997). If the
molecules are given sufficient mobility, they self-organize to minimize the free
energy of the material system arriving at one of several possible phase morpholo-
gies characterized by a small interaction area between the two unlike blocks.
Diblock copolymer molecules have been studied extensively in bulk, and complete
phase diagrams have been generated from both modeling and experiments. In bulk,
the phase morphology depends only upon the relative chain lengths of the two
blocks. A highly asymmetric diblock yields a spherical phase (spheres of the
smaller molecular weightblock A immersed in a matrix of block B) and an increas-
ing fraction of block A initiates the formation of a cylindrical phase, a gyroid phase,
and finally a lamellar phase is created for a symmetric diblock where A and B are
of similar chain lengths. Fig. 3.11 shows an example.
Spin Coating Methods
Spin coating is a simple but well-known technique to prepare thin films from liquid.
One of the most important applications of spin coating is to cast photoresistant
films for the lithography processes during microfabrication. A typical spin coating
40 Y. Zhao
(a)
A Block A B B
A B B
A A B
A B
A
A
Block B B
(b)
lamellae cylinders spheres
Fig. 3.10 (a) A block copolymer consists of homogeneous blocks, A (lighter) and B (darker),
connected by a covalent bond, drawn here for equal lengths of A and B blocks. (b) The resulting
morphology depends primarily on the relative length of the two blocks. Lamellae are typically
observed for similar lengths of A and B blocks. Shortening the B block produces morphologies
such as cylinders, and finally, for an even shorter B block, spheres (Harrison et al. 1997). (Reused
with permission from Journal of Vacuum Science and Technology B, 16, 544 (1998), Christopher
Harrison, Miri Park, Paul M. Chaikin, Richard A. Register, and Douglas H. Adamson, copyright
1998, AVS The Science and Technology Society)
Fig. 3.11 SEM micrographs of exposed and developed PS template for different annealing times.
Each sample was annealed undisturbed at 160°C under a vacuum for total time of (a) 3.5 h, (b) 6 h,
(c) 15 h, and (d) 34 h. Lower images have been filtered to enhance edges (Guarini et al. 2001).
(Reused with permission from Journal of Vacuum Science and Technology B, 19, 2784 (2001), K. W.
Guarini, C. T. Black, K. R. Milkove, and R. L. Sandstrom, copyright 2001, AVS The Science and
Technology Society)
Fig. 3.12 A schematic of spin coating process
process, as shown in Fig. 3.12, involves depositing a small volume of a liquid onto
the center of a substrate and then spinning the substrate at high speed (typically
>1000 rpm). The centripetal force during the rotation will cause the liquid to spread
to the edge of the substrate leaving a thin layer of liquid. After evaporation or
solidification occurs on the surface, this liquid layer becomes a thin film. Final film
thickness depends on the nature of the liquid such as the viscosity, drying rate,
percent solids, surface tension, and so on, and the spinning parameters. If the liquid
contains nanostructures, such as nanoparticles, colloids, nanorods, or even materi-
als that can form nanostructures, such as sol–gel liquids, block copolymers, and the
like, the resulting film after spin coating is a nanostructured thin film.
Electrochemical Deposition/Etching Methods
Electrochemical deposition has been widely used to deposit metal or semiconduc-

tor thin films. The substrate is placed into an electrolyte or a plating solution which
is an aqueous solution of a metallic salt. The electrolyte contains the positively
charged metallic ions due to the dissociation of the metallic salt. By passing a suf-
ficient amount of electric current through the electrolyte, the metal ions can be
reduced to solid metal on the substrate. This process is called electroplating or
electrochemical deposition. Electrochemical deposition is carried out in a three-
electrode electrochemical cell that contains a working electrode, a reference electrode,
and a counterelectrode. The working electrode must be conducting. Electrochemical
deposition is usually used for plating metallic or semiconducting materials into
nanometer-sized channels; that is, it is used in a so-called template-based method
discussed in the section, “Template-Based Synthesis Techniques.”
One unique electrochemical deposition method that has been recently developed
is called electrochemical atomic layer epitaxy or electrochemical atomic layer deposi-
tion (Stickney 2002). This method is developed by analogy with the atomic layer
deposition discussed in Section “Chemical Vapor Deposition Methods.” The surface-
limited adsorption in ALD is replaced by the underpotential deposition; during the
deposition of one element onto a second, frequently the first element will form an
42 Y. Zhao
Fig. 3.13 Nanochannels formed in anodized aluminum oxide films: the upper row shows the
results due to random nucleation (Li et al. 1998). (Reused with permission from Journal of
Applied Physics, 84, 6023 (1998), A. P. Li, F. Müller, A. Birner, K. Nielsch, and U. Gösele,
copyright 1998, American Institute of Physics.) The lower row results from regular molding on
the Al thin foil (Masuda et al. 1997). (Reused with permission from Applied Physics Letters, 71,
2770 (1997), Hideki Masuda, Haruki Yamada, Masahiro Satoh, Hidetaka Asoh, Masashi Nakao,
and Toshiaki Tamamura, copyright 1997, American Institute of Physics)
atomic layer at a potential less than that needed to deposit the element on itself. The
details of electrochemical ALD are discussed in Stickney (2002).
The opposite process for deposition is etching. During the process of elec-
trochemical etching of a metallic specimen, the reduction and oxidation process
(redox process) takes place. All metals in contact with the solution have a pro-
nounced tendency to become ionized by releasing (losing) electrons and
becoming oxidized. In electrolytic or anodic etching, an electrical potential is
applied to the specimen by means of an external circuit. A typical setup consists
of the specimen (anode) and its counterelectrode (cathode) immersed in an
electrolyte. Two well-known nanoporous thin films can be fabricated through
electrochemical etching: the anodized aluminum oxide (AAO) nanochannels,
and the porous Si structures. Fig. 3.13 shows the AAO fabricated through
anodization.
Electrospinning Methods
The electrospinning process is one of the major techniques to produce polymer

micro- or nanofibers (Huang et al. 2003). Recently, with the combination of
the sol–gel techniques, the metal oxide nanofibers can also be produced using the
same technique. A schematic diagram of the electrospinning setup is shown in
Fig. 3.14. The setup consists of three major components: a high-voltage power
Point Charge
1
Syringe
Pendant Drop
Envelope Cone
y0
Flat Ground Secondary Grounded Plate

−1
Collector Bending Instability x 0
(a) (b)
Fig. 3.14 (a) Schematic drawing of the electrospinning process with a flat ground collector. (b)
Numerical solution showing the equipotential lines and the direction of the electrostatic forces
(arrows) between a point charge and an infinite plate (Theron et al. 2001). (Reprinted by permis-
sion from Nanotechnology, 12, 384 (2001), copyright 2001, AIP)
Fig. 3.15 Left: SEM of electrospun type III collagen (human placenta) (Matthews et al. 2002).
(Reprinted by permission from Biomacromolecules 3, 232 (2002), copyright 2002, American
Chemical Society.) Right: SEM images of aligned nanofibers that were collected on a carbon tape
attached to the edge of the disc collector (Theron et al. 2001). (Reprinted by permission from
Nanotechnology, 12, 384 (2001), copyright 2001, AIP )
supply, a small-diameter capillary tube that connects to a syringe pump, and

a metal collection plate/screen/drum. During the electrospinning process, a
polymer solution, melt, or unsolidified sol–gel solution with proper viscosity is
44 Y. Zhao
loaded into a syringe. A high voltage is created between the capillary tube and
the collection plate so that an electrically charged liquid jet is generated from the
end of the capillary tube. Before reaching the collection plate, the liquid jet
evaporates or solidifies and is collected as an interconnected web on the plate.
The critical parameters for electrospinning are voltage, the distance between the
collection plate and the capillary tube, the properties of the liquid, such as vis-
cosity, evaporation, or solidification rate, syringe pumping speed, and so on.
Two examples are shown in Fig. 3.15. With the process conditions properly
tuned along with the proper designing of the collection method, capillary tubes,
porous nanofibers, aligned nanofibers, tubelike fibers, and so on, are all
achievable.
Template-Based Synthesis Techniques
The template-based synthesis technique is a simple and straightforward method

to fabricate NSTFs. The basic idea behind this method is to treat the grown
nanostructures as scaffolds or molds, and apply other fabrication or synthesis
techniques to completely or partially replicate their structures. The template
can be any nanostructure including nanoparticles, nanorods, nanowires, nano-
porous materials, colloids, and even DNA or proteins. This method involves at
least two major steps: obtaining the template, and replicating the template.
The fabrication of the template can be performed by one of the synthesis tech-
niques introduced in the above sections whereas the replication can be done by
another fabrication technique. Thus, the template-based synthesis technique
will be a combination of at least two different synthesis techniques. A detailed
review of the template-based methods can be found in Xia et al. (2003), Martin
(1994), and Huczko (2000).
Direct Replication Methods
The general processing steps for direct replication are shown in Fig. 3.16. The
template is a nanoporous material, such as anodized alumina. Electrochemical
plating is usually used to fill the channels. Thus, a back electrode is first depos-
ited by thermal evaporation of Au or other metals. With the help of this elec-
trode, other materials can be electrochemically plated into the channels. After
chemically dissolving the template material, a nanostructure that replicates the
template is formed. Many metallic nanorod structures have been fabricated by
this method. Note that during the electrochemical plating, one could alterna-
tively change the materials plated into the channel. Thus, multilayered nanorod
structures can be fabricated.
(a) Nanoporous template (b) Back electrode coating (c) Electroplating desired materials (d) Disolving the template
Fig. 3.16 The direct replication process for template-based synthesis technique
(a) Nanorod template (b) Surface coating of nanorods (c) Removing the nanorod template
Fig. 3.17 A typical partial replication process: using nanorods as template to generate nano-
tubes
Partial Replication Methods
On the contrary, the partial replication method will not fill the channel fully. Rather,
as shown in an example of Fig. 3.17, the channel will be partially filled. In Fig.
3.17, the template is an aligned nanorod array. By using a CVD or ALD technique,
a uniform layer of desired material can be conformally coated on the surface of the
nanorods with a controllable thickness. This kind of transition structure can
be called a core-shell structure. After removing the top layer through a dry etching
process such as plasma etching or reactive ion etching, and chemically dissolving
the nanorod template, an array of nanotubes of the desired material will form.
However, the template is not limited to the nanorods, nanoparticles, such colloids,
or nanowires, and can serve as such a template to make either core-shell structures
or shelllike or tabulatelike structures.
Glancing Angle Deposition (GLAD)
The glancing angle deposition technique is based on a simple modification of the

deposition configuration of a physical vapor deposition system (Robbie and Brett
1997). Any thin film physical vapor deposition system such as thermal evaporation,
46 Y. Zhao
sputtering growth, or pulsed laser deposition, and the like, can be readily converted
into a GLAD system. In fact, the GLAD technique is the extension of the com-
monly used oblique angle deposition (OAD) which has been practiced for many
years by the thin-film deposition community. Most of its fundamental growth
mechanisms are similar to the oblique angle deposition although deviation may be
expected.
Oblique Angle Deposition (OAD)
The experimental setup for oblique angle deposition is shown in Fig. 3.18a. The
collimated evaporation beam has an incident angle q (³70° in general) with respect
to the substrate surface normal. The incoming vapor flux can be treated as a vector
F as shown in Fig. 3.18b. The flux has two components: a vertical component F^ =
F cosq and a lateral component (a vector) F|| with | F | = F sinq. The substrate will
||
receive the vapor flux from both vertical and lateral directions.
During the deposition onto a flat substrate, the impinging atoms will randomly
form islands on the substrate as shown initially in Fig. 3.19a. As deposition pro-
ceeds, the nucleated islands will act as further growth centers, and all the tall
islands will receive more impinging atoms as compared to the shorter ones due to
a so-called shadowing effect. This competition process will only leave the tallest
islands to grow into columns and a nanocolumnar film will be formed (Fig. 3.19b).
The lateral component F|| is the source for the shadowing effect.
For oblique angle deposition, F|| remains constant during the deposition, and a
columnar film with tilt angle b will be formed. Fig. 3.20 shows the cross-sectional
SEM images of Si thin films deposited at different incident angles. At q = 0°, a
continuous and uniform thin film is formed; at q = 30°, small columns begin to
grow; at q = 60°, the columnar structure becomes more obvious; at q = 80°, obvi-
ous columnar structures are formed into a film. In general, the column tilt angle b
is less than the vapor incident angle q which follows the empirical tangent rule,
Fig. 3.18 (a) Experimental setup for oblique angle deposition. (b) The incident flux F can be
decomposed into two different components, F⊥ is the flux perpendicular to substrate, and F|| is the
flux parallel to the substrate
Fig. 3.19 The shadowing effect during oblique angle deposition: (a) initial nucleation to form
shadowing centers, and (b) columnar structures formed due to the shadowing effect
Fig. 3.20 SEM cross-section view of Si films fabricated at different incident angles. The scale
bars are 100 nm
tan b = 1/2 tan q for small q (Nieuwenhuizen and Haanstra 1996; Lisfi and Lodder
1 − cos q ⎞
2001), or the cosine rule, b = q − arcsin ⎛⎜ ⎟ (Trait et al. 1993).
⎝ 2 ⎠
48 Y. Zhao
Fig. 3.21 Multilayer Fe/Ni structure deposited by OAD: (a) SEM top view, and (b) SEM cross-
section view
Thin films deposited by oblique angle deposition have the following character-
istics: the films are in general porous and they possess nanocolumnar structures.
The nanocolumns tilt away from the surface normal to the incident evaporation
beam direction. The size and density of the nanocolumns change as a function of
the incident angle q (Fig. 3.20). Furthermore, if one changes the deposition source
with a fixed deposition configuration as shown in Fig. 3.1, multilayered columnar
films can also be grown. Fig. 3.21 shows the SEM top view and cross-sectional
view of a Ni/Fe two-layer nanocolumnar film deposited by OAD. This structure
also shows the thermoelectric effect upon heating.
Glancing Angle Deposition (GLAD)
Although the nanocolumnar films generated by oblique angle deposition can be

treated as one kind of NSTF, the morphology of the structures is not easy to control
during the growth. The preferred tilting columnar angle b introduces anisotropy in
the films which is not desirable in most cases. In order to better control the orienta-
tion as well as the structure of the nanocolumns, the glancing angle deposition
technique is developed as shown in Fig. 3.22.
The basic deposition setup is exactly the same as that of oblique angle deposi-
tion; the only difference is that the substrate is manipulated by two stepper motors:
one motor controls the incident angle q, and the other motor controls the azimuthal
rotation of the substrate with respect to the substrate surface normal. During the
deposition, the substrate can rotate azimuthally at a fixed incident angle or rotate
back and forth changing the incident angle or rotate azimuthally and polarly simul-
taneously. The movement of the two motors is controlled by a computer. By chang-
ing the speed and phase of the azimuthal rotation and/or the polar rotation, the
(a) (b)
F
Fig. 3.22 (a) Experimental setup for GLAD; (b) cancellation of the F|| term due to the rotation.
Only F^ contributes effectively to the increasing height of the nanocolumns
combination of the two rotations along with manipulation of the deposition rate, the
nanocolumns can be sculptured into a C-shape, S-shape, zigzag shape, matchstick,
helical, or vertical columns (Robbie and Brett 1997; Young and Kowal 1959;
Motohiro and Taga 1989; Azzam 1992; Robbie et al. 1996; Messier et al. 1997,
2000; Liu et al. 1999; Malac and Egerton 2001; Malac et al. 1999; Zhao
et al. 2002a, b; Ye et al. 2002). In fact, the most intriguing aspect of GLAD is that
the structures of the nanorods can be well designed by programming the substrate
rotations. The following gives examples of various nanostructures fabricated in
such a way.
Growth of Vertically Aligned Nanorod Arrays
The simplest configuration for GLAD is to rotate a substrate azimuthally with a

constant speed while fixing the deposition angle q. Depending on the rotation speed
w and the deposition rate R, both vertically aligned nanocolumns and helical nanos-
tructures can be formed. The programmed azimuthal rotation is shown in Fig.
3.23a, and the steady slope represents a constant rotation speed. Figs. 3.23b and c
show an example of a Si nanocolumnar structure deposited with a constant azi-
muthal rotation speed of 0.0019 rev/s and 100 total substrate revolutions. The depo-
sition angle q was fixed at 85°, and the growth rate R was 0.2 nm/s. The nanocolumns
are all aligned vertically with respect to the substrate, and the locations of the col-
umns are random due to random nucleation on a flat Si substrate.
The formation of the vertically aligned nanocolumns is expected if we consider
the direction of the flux as shown in Fig. 3.18b. Because the substrate rotates azi-
muthally, it has an equal chance to receive the same amount of vapor within the
same azimuthal angular sector because F|| is a constant at a fixed q. After a com-
plete revolution, the average flux parallel to the substrate surface, SF||, is zero due
50 Y. Zhao
Fig. 3.23 (a) The program sequence for vertically aligned nanocolumn growth, (b) SEM top
view, and (c) SEM cross-sectional view of the Si nanorods grown by continuous rotation. The
scale bars are 200 nm (Zhao et al. 2002a)
to the cancellation of F|| at opposite directions, whereas the effective F⊥ is a con-

stant. Thus, there is no preferred growth orientation of the nanocolumn as shown in
Fig. 3.22b. However, at a specific time, the growth is still an oblique deposition, and
the shadowing effect still remains as a major mechanism to control the growth. The
purpose of the azimuthal rotation is to constantly adjust the column tilting direction
to make nanocolumns straight.
Growth of Helical Nanostructures
If the substrate rotation speed is intentionally slowed at each rotation interval, there
will be enough vapors deposited onto the substrate to form a column with the pre-
ferred orientation. The result of this slow rotation gives a continuous formation of
nanocolumns on top of each other and along different directions, that is, the forma-
tion of helical nanorod structures (Zhao et al. 2002a). By controlling the time
interval between each rotation step as well as the incident angle, one can control the
length, diameter of the pitch, diameter of the rod, and the number of turns of the
helical nanorod structure. For example, to fabricate square spiral posts (or square
spirals) with a circumference of L for one pitch, we fix the flux incident angle at
85° and program the stepper motor with the following sequence.
1. Grow a layer with a thickness of L/8 without rotating the substrate.
2. Rotate 90° at a fast azimuthal rotation rate.
3. Repeat steps 1 and 2.
4. Grow a layer with a thickness of L/4 without rotating the substrate.
5. Rotate 90° at a fast azimuthal rotation rate.
6. Repeat steps 4 and 5 to the desired pitch number for the square spirals.
The expected square spiral structure is shown in Fig. 3.24a. The Si square spiral
arrays on bare Si substrates fabricated through this method are shown in Figs. 3.24b
and c (Zhao et al. 2002a). Both the top-view and cross-sectional SEM images show
that the spirals are uniformly distributed across the whole surface with almost the
Fig. 3.24 (a) The sequence for fabricating a two-turn square spiral (see the detailed description
in the text), and SEM images of two-turn square Si spirals; (b) top-view; and (c) cross-section of
Si spirals on a bare Si substrate. The arrows on (c) indicate the growth sequence in a spiral, which
is similar to what we expect in (a) (Zhao et al. 2002a)
Fig. 3.25 (a) The program sequence for square nanospirals, and (b) the SEM cross-sectional
image of ten-turn square Si spirals. Due to short pitch, the nanospirals appear well separated from
each other
same length. However, each spiral is closely packed to the adjacent spiral which
means that the whole film should act as a spiral bed instead of individual spirals. The
structure of the spiral is described by the arrows in Fig. 3.24c. From the substrate to
the top of the spiral, the arrow changes direction ten times; the first two direction
changes are small, which correspond to the L/8 arms and the remaining eight turns
correspond to the eight L/4 arms representing two complete turns of the square spi-
ral. This structure is similar to what we expect in Fig. 3.24a. The diameter of the Si
arm is about 50 nm, and the diameter of the spiral is about 200–500 nm. The twist
angle of the spiral is determined by the incident flux angle because the fabrication
process is mostly under a fixed angle deposition. The resulting tilt angle (similar to
the experiment shown in Fig. 3.20 was determined to be 55° ± 2°. By adjusting the
L parameter, smaller-sized spiral nanostructures can be formed. Fig. 3.25 shows the
programmed azimuthal rotation for square spiral structures and an example of a ten-
turn spiral film. The deposition conditions are almost exactly the same as the previ-
ous deposition except for the value of L = 200 nm.
52 Y. Zhao
Fig. 3.26 The program sequence examples for (a) zigzag, and (b) C-shape nanorods
Fig. 3.27 The SEM (a) top-view, and (b) cross-section images of multilayer Si spiral/straight
nancolumns grown on a bare Si(100) substrate. There are two layers of six-turn spirals and two
layers of 300 nm long columns (Zhao et al. 2002a)
Other rotation programming can be established to form other nanorod structures,

such as zigzag, C-shape, or S-shape structures. The rotation programming within a
revolution required to fabricate zigzag and C-shape structures is shown in Fig. 3.26.
Clearly, other helical structures can also be fabricated by changing the program-
ming time sequences.
Integration of Multilayered Nanorod Structures

with Different Morphology
Because different azimuthal rotation programming produces different nanorod

structures by more than two different programming sequences discussed above,
multilayered nanostructures with different topographies can be fabricated. Fig. 3.27
shows one example of the multilayer square spiral/straight Si nanorods (Zhao et al.
2002a). To fabricate the multilayer square spiral/straight nanocolumns, one must
first fabricate the spiral arrays to a preset pitch and the number of turns according
Fig. 3.28 The SEM cross-section images of double handed TiO2 triangular spirals (van Popta
et al. 2005). (Reused with permission from Journal of Applied Physics, 98, 083517 (2005), Andy
C. Van Popta, copyright 2005, American Institute of Physics)
to the program shown in Fig. 3.25a, then continuously rotate the substrate at a faster
speed to build the straight nanocolumns according to the program shown in Fig.
3.23a. One would then repeat the whole process. For the film deposited in Fig. 3.27,
the growth conditions are R = 0.32 nm/s, L = 120 nm, and the rotational speed for
the columns is 0.0019 rev/s. There are two layers of six-turn spirals and two layers
of 300 nm long columns. The two spiral sections and two column sections are
almost the same. By changing the computer program in a similar manner, other
kinds of multilayered nanostructures can be fabricated. For example, by changing
the rotation directions, van Popta et al. (2005) have fabricated double-handed trian-
gular helical structures from TiO2 as shown in Fig. 3.28.
Integration of Multilayered Nanorod Structures

with Different Materials
In the same spirit, multilayered nanostructures with different materials can be fab-
ricated using a multilayer deposition system; that is, the source can be changed
during the deposition. Two simple examples are shown in Fig. 3.29: one is a match-
sticklike Cu/Si nanorod structure, and the other one is a Ni rod/Si spiral structure.
The morphology of these two structures is relatively simple since one material
forms a layer of simple morphology. However, by programming the substrate rota-
tion and the source material simultaneously, one can fabricate even more compli-
54 Y. Zhao
Fig. 3.29 The SEM cross-section images of (a) Cu/Si matchsticklike nanorod structure, and (b)
Ni rod/Si spiral structure
Fig. 3.30 The program sequence for growth of Ni/Si multilayer nanospiral structure and expected
morphology
cated structures. Fig. 3.30 shows the program sequence to make a Si/Ni multilayer
spiral (He et al. 2007):
1. Grow a layer of Si nanorods of 800 nm in deposition thickness without rotating
the substrate.
2. Close the source shutter, rotate the substrate 90°; at the same time, replace the
Si source by a Ni source.
3. Grow a layer of Ni nanorods of 800 nm in deposition thickness without rotating
the substrate.
4. Close the source shutter, and rotate the substrate 90°; at the same time, replace
the Ni source by the Si source.
5. Repeat steps 1, 2, 3, and 4 to obtain the desired two-turn and eight-arm square
nanosprings.
Fig. 3.31 The SEM (a) top view, and (b) cross-section images of Ni/Si multilayer nanospiral
structure. The arrows on (b) indicate the growth sequence in a spiral. (c) The TEM image of a
bundle of Ni/Si multilayer nanospirals. The dark parts represent Ni and light parts are Si
(He et al. 2007)
The flux incident angle was fixed at 86°, and the substrate was rotated azimuth-
ally in the same direction. The resulting Si/Ni multilayer spiral structures are shown
in Fig. 3.31. The SEM top view and cross-sectional view show that the nanorods
are similar to the square spiral as shown in Fig. 3.24, but the TEM image shows
different contrasts due to different materials; the dark parts are Ni nanorods, and the
lighter parts are Si nanorods.
Phase Modulation: Controlling the Shape of the Nanorods
For the nanostructure fabrication described in the previous section, we have only
utilized either a continuous rotation or interrupted rotation. However, if we intro-
duce two different rotation speeds during one evolution, then the relative phase
difference between these two rotation speeds can be used to tune the topology of
the nanostructures (Zhao et al. 2002b; Ye et al. 2002). One obvious way to incor-
porate the phase factor is to introduce a symmetric rotation with two different rota-
tion speeds. In order to do so, we divide each revolution during deposition into 2N
phase sectors where N represents the N-fold symmetry of the desired shape of the
columns. The rotation speed was alternately changed between wl and wh (wh>> wl)
for adjacent sectors. Because wh is much higher than wl, within the sector with
rotational speed wh, the sample will receive a smaller amount of vapor per radian
than that with wl. Thus, the shape of the nanorods can be tuned by changing the
ratio of wh/wl as well as the phase sector N.
Fig. 3.32 shows a program sequence for a twofold symmetry rotation GLAD.
Different slopes in different phase sectors indicate the difference in rotation speeds.
Fig. 3.33 shows the resulting Si nanorod structures with two-, three-, and fourfold
symmetry rotation (Zhao et al. 2002b). The examples show that different symmetrical
rotations alter the lateral arrangement of the nanorod arrays while also changing the
shape or aggregation of the nanorods. Due to rotational symmetry, all the nanorods
are aligned with each other and are perpendicular to the substrate surface.
56 Y. Zhao
Fig. 3.32 The program sequence for twofold symmetry rotation
Fig. 3.33 The SEM top view images of different GLAD samples with different rotational sym-
metries. The upper row lists films deposited on bare Si(100) substrates, and the lower row lists
films deposited on the colloid substrates which have point defects (Zhao et al. 2002b)
Phase Modulation: Controlling the Orientation of the Nanorods
If the phase sectors are not divided symmetrically, different sectors will receive
different amounts of vapor; one should expect that the tilting angle of the nanorods
can be modulated (Ye et al. 2002). One simple case is to divide one revolution into
two different phase sectors as shown in Fig. 3.34. In one sector, the substrate rotates
Azimuthal Rotation Angle

2π
0
0 T/2 T
Program Time
Fig. 3.34 The program sequence for asymmetry rotation
at speed wl for an angular range of f, and in the remaining sector, the substrate
rotates at speed wh for an angular range of 2p – f . Because the asymmetry in the
f direction is intentionally introduced, there will be a net F|| after a complete revo-
lution, and there will be an effective vapor incident angle a different from q. The
apparent lateral flux F¢|| will be changed according to the sizes of sectors and the
ratio of the rotational speeds:
⎡ f / 2 F|| cos ϕ −f / 2
F|| cos ϕ ⎤ 2 F|| ⎛ 1 1⎞ f
F¢ || = ⎢ ∫ dϕ + ∫ dϕ ⎥ / t = ⎜ − ⎟ sin . (3.1)
⎢⎣ −f / 2 w l f /2
wh ⎥⎦ t ⎝ wl wh ⎠ 2
The total time t for a complete revolution is:
f 2p − f (3.2)
t= + .
wl wh
Therefore the effective flux incident angle a can be expressed by
⎛ 1 1⎞ f ⎛ wh ⎞
2 F sin q ⎜ − ⎟ sin ⎜⎝ w − 1⎟⎠
F¢ || ⎝ wl wh ⎠ 2 f (3.3)
tan a = = = 2 tan q sin l
.
F⊥ ⎛ f 2p − f ⎞ 2 ⎛w ⎞
F cos q ⎜ + 2p + ⎜ h − 1⎟ f
⎝ wl w h ⎟⎠ ⎝ wl ⎠
Because the experimental incident angle q is fixed, the column density should
be constant whereas the effective incident angle a is changed by changing f and
wh/wl. We expect the tilt angle b will change continuously; the continuous tuning of
58 Y. Zhao
the tilt angle has been demonstrated experimentally (Ye et al. 2002). A special case
is if the substrate is rotated uniformly, wh/wl = 1; then tan a= 0. The deposited
columns have no preferential deposition angle, and the columns stand up along the
normal of a substrate.
Formation of Regular Array of Nanorods by Templates
Almost all the nanocolumnar films shown above are deposited onto bare flat Si
substrates. In these cases, the random nucleation is the deterministic mechanism for
the lateral distribution of the nanocolumns. However, if we can use one of the major
growth mechanisms of GLAD, the shadowing effect, properly, we can grow a high
aspect ratio regular array of nanocolumns. To achieve this goal, we can use a sub-
strate with a proper two-dimensional nanotemplate as shown in Fig. 3.35a. The
features on the template will act as shadowing centers, and the deposition particles
will only accumulate onto the shadowing centers under the proper geometric depo-
sition conditions especially when the vapor incident angle q is larger than the critical
angle qc. The critical angle qc, as shown in Fig. 3.35a, is determined by the geo-
metry of the template, tan q c = L / h . Fig. 3.35b also shows an example of a regular
helical Si spring grown onto regular W plugs fabricated by photolithography. Clearly,
the use of a template provides a lateral control of the nanocolumnar growth.
Control of NSTF Film Porosity
All the above fabrication methods concentrate on a single degree of freedom for the
substrate manipulation which is azimuthal rotation. In fact, there are two other
rotational degrees of freedom one can utilize: the polar rotation to change the
particle incident angle q, and the nutation to rotate the substrate to face a different
(a)
D L q<qc
q>qc
qc
h
Fig. 3.35 (a) The regular array template and the geometry of the GLAD: the definition of the
critical incident angle qc; (b) SEM cross-section view of regular Si spirals grown onto regular W
plugs fabricated by photolithography method. The scale bar is 1 mm
Fig. 3.36 (a) The SEM cross-section view of three-dimensional porous Si structure fabricated by
combining azimuthal and polar rotations (detailed sequence is described in the text). (b) A cap-
ping layer formed on top of the nanorod surface (Robbie and Brett 1997). (Reused with permis-
sion from Journal of Vacuum Science and Technology A, 15, 1460 (1997), K. Robbie and M. J.
Brett, copyright 1997, AVS The Science and Technology Society)
phase of the incident flux. The polar rotation has been demonstrated to change the
size and density of the nanocolumns as shown in Fig. 3.20 and the nutation has not
been used for nanocolumn fabrication.
In principle, one can control all three rotational degrees of freedom for rotation
to fabricate the desired nanostructures. Fig. 3.36a shows an example of a porous Si
thin film fabricated by combining azimuthal rotation and polar rotation. To fabri-
cate this film, the sequence of the deposition is as follows.
1. Open shutter.
2. Set q = 85°, and deposit for 200 s.
3. Change q to 75°, and deposit for 100 s.
4. Change angle q to 20°, and deposit for 100 s.
5. Close shutter.
6. Change angle q from 20° to 85°.
7. Repeat steps 1–6 for another six times (seven layers).
During the deposition, the azimuthal rotation proceeded with a rotation speed of
0.5 rev/s, and the deposition rate was 0.5 nm/s. Other similar structures can be
designed through the two motor controls. For example, by programming the inci-
dent angle, one can put a capping layer onto the nanorod structures (Fig. 3.36b;
Robbie and Brett 1997).
Formation of Nanostructured Architectures on Optical Fiber
By slightly changing the configuration of the GLAD setup, we can deposit aligned
nanostructures onto cylindrical optical fibers (Fan and Zhao 2005). The experimental
setup is shown in Fig. 3.37. Inside a PVD chamber, a cylindrical object such as an
optical fiber is installed coaxially onto a stepper motor (Motor2) through a pin vise.
60 Y. Zhao
Motor2
Motor1 Azimuthal
Polar rotation
rotation
Rod mount:
pin vise, etc.
Uniform rod
q
Evaporation
source
Fig. 3.37 Experimental setup for nanorod deposition onto a cylindrical object (Fan and Zhao
2005)
The object can be rotated in both polar (by Stepper Motor1) and azimuthal (by
Stepper Motor2) directions controlled by a computer. In order to deposit aligned
nanorods onto the surface of a cylindrical object, the axis of the object is slightly
rotated to a polar angle q < 10° by Motor1. The source material is evaporated onto
the object at a constant rate. Therefore, during the deposition the incoming vapors
can be shadowed by nanostructures (including initial islands) on the surface, and
the cylindrical object is rotated at a constant speed about its axis by Motor2 to
expose the entire surface to the vapor in order to achieve uniform coating.
Fig. 3.38 shows the Ag nanorod deposited onto a silica optical fiber with core
diameter of 98.6 m. Other materials such as Cu, Si, Ni, Al, and TiO2 have also been
fabricated as nanorods onto optical fibers using this same method. These prelimi-
nary results show that it is possible to integrate well-aligned nanorod structures onto
optical fiber-based sensors; therefore, a nanostructure-based high-sensitivity sensor is
possible. With a slight modification of the deposition configuration, one can also
put nanorod arrays onto tapered objects. For example, Fig. 3.39a shows an Ag
nanorod array on a tapered micropipette surface. With programming both Motor1
and Motor2, one can also deposit multilayer film/nanorod (Fig. 3.39b) and nano-
rod/nanorod (Fig. 3.39c) structures onto optical fibers. These results demonstrate
that it is feasible to directly incorporate complicated nanoscale architectures onto
optical fibers. The nanorods can also be deposited onto the tip of an optical fiber.
Fig. 3.38 Aligned Ag nanorods on an optical fiber (Fan and Zhao 2005)
Fig. 3.39 (a) Ag nanorods on tapered micropipette tip; (b) Ag film/nanorod two layered struc-
ture; and (c) Ag nanorod/Si nanorod multilayered structure on an optical fiber. The arrows point
to the Ag/Si interface (Fan and Zhao 2005)
Fig. 3.40 The experimental configuration for depositing nanorods on the cross-section of an
optical fiber and an example of Ag nanorods on the core of an optical fiber (Fan and Zhao 2005)
Fig. 3.40 shows the deposition geometry and the resulting Ag nanorods on the core
of the surface. These results demonstrate that it is feasible to directly incorporate
complicated nanoscale architectures onto optical fibers using GLAD.
62 Y. Zhao
Summary on GLAD
We use numerous examples to demonstrate that the GLAD technique has the fol-
lowing advantages in terms of fabricating NSTFs. (1) It can form vertical aligned
nanorod arrays. (2) The size and density of the nanorods can be controlled by the
vapor incident angle. (3) There is virtually no materials limit. As long as the mate-
rial can be evaporated, it can form vertical aligned nanorod structures. (4) The
shape, alignment, and orientation of the nanorods can be easily changed by pro-
gramming the rotation procedures. (5) The shadowing effect introduces a self-
alignment effect where templates can be used to mediate the size and orientation of
the nanorods. (6) Three-dimensional-shaped nanorod structures can be sculptured
by computer programming. (7) Multilayered nanostructures are achievable using a
multilayer deposition system. (8) It is a low-temperature PVD process, and it is
compatible with microfabrication processes. For example, Brett et al. have inte-
grated nanostructures fabricated by GLAD into microfluidic devices (Kiema et al.
2005). The most intriguing fact of GLAD is that the structures of the nanorods can
be well-designed in such a way. This cannot be achieved by other nanostructure
fabrication techniques. However, the potential of GLAD techniques for device fab-
rication has yet to be fully explored.
References
Azzam RMA (1992) Chiral thin solid films: Method of deposition and applications. Appl. Phys.
Lett., 61:3118–3120.
Bartlett PN, Baumberg JJ, Coyle S, Abdelsalam ME (2004) Optical properties of nanostructured
metal films. Faraday Discuss., 125:117–132.
Biswas A, Aktas OC, Schürmann U, Saeed U, Zaporojtchenko V, Faupel F, Strunskus T (2004)
Tunable multiple plasmon resonance wavelengths response from multicomponent polymer-
metal nanocomposite systems. Appl. Phys. Lett., 84:2655–2657.
Biswas A, Marton Z, Kanzow J, Kruse J, Zaporojtchenko V, Faupel F, Strunskus T (2003)
Controlled generation of Ni nanoparticles in the capping layers of Teflon AF by vapor-phase
tandem evaporation. Nano Lett., 3:69–73.
Caruso RA, Antonietti M (2001) Sol-gel nanocoating: An approach to the preparation of structured
materials. Chem. Mater., 13:3272–3282.
Fan JG, Zhao YP (2005) Direct deposition of aligned nanorod array onto cylindrical objects.
J. Vac. Sci. Technol. B, 23:947–953.
Guarini KW, Black CT, Milkove KR, Sandstrom RL (2001) Nanoscale patterning using self-
assembled polymers for semiconductor Applications. J. Vac. Sci. Technol. B, 19:2784–2788.
Harrison C, Park M, Chaikin PM, Register RA, Adamson DH (1997) Lithography with a mask of
block copolymer microstructures. J. Vac. Sci. Technol. B, 16:544–552.
He YP, Fu JX, Zhang Y, Zhao YP, Zhang LJ, Xia AL, Cai JW (2007) Multilayered Si/Ni nano-
springs and their magnetic properties. Small, 3:153–160.
Huang MH, Wu YY, Feick H, Tran N, Weber E, Yang PD (2001) Catalytic growth of zinc oxide
nanowires by vapor transport. Adv. Mater., 13:113–116.
Huang ZM, Zhang YZ, Kotaki M, Ramakrishna S (2003) A review on polymer nanofibers by
electrospinning and their applications in nanocomposites. Composites Sci. Technol.,
63:2223–2253.
Huczko A (2000) Template-based synthesis of nanomaterials. Appl. Phys. A, 70:365–376.

Kiema GK, Jensen MO, Brett MJ (2005) Glancing angle deposition thin film microstructures for
microfluidic applications. Chem. Mater., 17:4046–4048.
Kim H (2003) Atomic layer deposition of metal and nitride thin films: Current research efforts and
applications for semiconductor device processing. J. Vac. Sci. Technol. B, 21:2231–2261.
Levitt AP (ed.) (1970) Whisker Technology. Wiley, New York.
Li AP, Müller F, Birner A, Nielsch K, Gösele U (1998) Hexagonal pore arrays with a 50–420 nm
interpore distance formed by self-organization in anodic alumina. J. Appl. Phys.,
84:6023–6026.
Lieber CM (1998) One-dimensional nanostructures: Chemistry, physics and applications. Solid
State Commun., 107:607–616.
Lisfi A, Lodder JC (2001) Magnetic domains in Co thin films obliquely sputtered on a polymer
substrate. Phys. Rev. B, 63:174441.
Liu F, Umlor MT, Shen L, Weston J, Eads W, Barnard JA, Mankey GJ (1999) The growth of
nanoscale structured iron films by glancing angle deposition. J. Appl. Phys., 85:5486–5488.
Lyu SC, Zhang Y, Ruh H, Lee HJ, Shim HW, Suh EK, Lee CJ (2002) Low temperature growth and
photoluminescence of well-aligned zinc oxide nanowires. Chem. Phys. Lett., 363:134–138.
Malac M, Egerton RF (2001) Observations of the microscopic growth mechanism of pillars and
helices formed by glancing-angle thin-film deposition. J. Vac. Sci. Technol., A19:158–166.
Malac M, Egerton RF, Brett MJ, Dick B (1999) Fabrication of submicrometer regular arrays of
pillars and helices. J. Vac. Sci. Technol. B, 17:2671–2674.
Martin CR (1994) Nanomaterials: A membrane-based synthesis approach. Science, 266:1961–1966.
Masuda H, Yamada H, Satoh M, Asoh H, Nakao M, Tamamura T (1997) Highly ordered nano-
channel array architecture in anodic alumina. Appl. Phys. Lett., 71:2770–2772.
Matthews JA, Wnek GE, Simpson DG, Bowlin GL (2002) Electrospinning of collagen nanofibers.
Biomacromolecules, 3:232–238.
Messier R, Gehrke T, Frankel C, Venugopal VC, Otaño W, Lakhtakia A (1997) Engineered sculp-
tured nematic thin films. J. Vac. Sci. Technol. A, 15:2148–2152.
Messier R, Venugopal VC, Sunal PD (2000) Origin and evolution of sculptured thin films. J. Vac.
Sci. Technol. A, 18:1538–1545.
Motohiro T, Taga Y (1989) Thin film retardation plate by oblique deposition. Appl. Opt.,
28:2466–2482.
Ng HT, Chen B, Li J, Han J, Meyyappan M, Wu J, Li SX, Haller EE (2003) Optical properties of
single-crystalline ZnO nanowires on m-sapphire. Appl. Phys. Lett., 82:2023–2025.
Nieuwenhuizen JM, Haanstra HB (1966) Microfractography of thin films. Philips Tech. Rev.,
27:87–89.
Pomogailo AD (2005) Polymer sol-gel synthesis of hybrid nanocomposites. Colloid J.,
67:658–677.
Robbie K, Brett MJ (1997) Sculptured thin films and glancing angle deposition: Growth mecha-
nisms and applications. J. Vac. Sci. Technol. A, 15:1460–1665.
Robbie K, Brett MJ, Lakhtakia A (1996) Chiral sculptured thin films. Nature, 384:616–616.
Stickney JL (2002) Electrochemical atomic layer epitaxy (EC-ALE): Nanoscale control in the
electrodeposition of compound semiconductors. In: Alkire RC, Kolb DM (eds.) Advances in
Electrochemical Science and Engineering, Volume 7. Wiley-VCH, Weinheim, Germany.
Theron A, Zussman E, Yarin AL (2001) Electrostatic field-assisted alignment of electrospun
nanofibres. Nanotechnology, 12:384–390.
Trait RN, Smy T, Brett MJ (1993) Modeling and characterization of columnar growth in evapo-
rated-films. Thin Solid Films, 226:196–201.
van Popta AC, Brett MJ, Sit JC (2005) Double-handed circular Bragg phenomena in polygonal
helix thin films. J. Appl. Phys., 98:083517.
Wang ZL (2004) Functional oxide nanobelts: Materials, properties and potential applications in
nanosystems and biotechnology. Annu. Rev. Phys. Chem., 55:159–196.
Wei BQ, Vajtai R, Jung Y, Ward J, Zhang Y, Ajayan PM, Ramanath G (2002) Organized assembly
of carbon nanotubes. Nature, 416:495–496.
64 Y. Zhao
Wu YY, Yan HQ, Huang M, Messer B, Song JH, Yang PD (2002) Inorganic semiconductor
nanowires: Rational growth, assembly, and novel properties. Chem. Eur. J., 8:1260–1268.
Xia YN, Yang PD, Sun YG, Wu YY, Mayers B, Gates B, Yin YD, Kim F, Yan HQ (2003) One-
dimensional nanostructures: Synthesis, characterization, and applications. Adv. Mater.,
15:353–389.
Yang PD, Yan HQ, Mao S, Russo R, Johnson J, Saykally R, Morris N, Pham J, He RG, Choi HJ
(2002) Controlled growth of ZnO nanowires and their optical properties. Adv. Funct. Mater.,
12:323–331.
Ye DX, Zhao YP, Yang GR, Zhao YG, Wang GC, Lu TM (2002) Manipulating the column tilt
angles of nanocolumnar films by glancing angle deposition. Nanotechnology, 13:615–618.
Young NO, Kowal J (1959) Optically active fluorite films. Nature, 183:104–105.
Zhao YP, Ye DX, Wang GC, Lu TM (2002b) Novel nano-column and nano-flower arrays by
glancing angle deposition. Nano Lett., 2:351–354.
Zhao YP, Ye DX, Wang PI, Wang GC, Lu TM (2002a) Fabrication Si nano-columns and square
springs on self-assembly colloid substrates. Int. J. Nanosci., 1:87–97.
Chapter 4
Integrated Micromachining Technologies
for Transducer Fabrication
Wei-Cheng Tian
Abstract In order to design a microfabricated transducer, the full understanding of

various micromachining technologies is essential. The manufacturability of trans-
ducer structures has to be considered and the sensing material application has to be
compatible with the transducer fabrication. Various technologies, such as lithography,
pattern transfer, and platform material choices are discussed first followed by
reviews of different sensing platforms and sensitive material integration techniques.
This chapter presents a summary of start-of-the-art micromachining technologies
for transducers.
Introduction
In general, a micromachined sensing element consists of two major parts: the trans-
ducer and the sensing material. Various micro- and nanotechnologies can be used to
fabricate the sensing element. These micro- and nanotechnologies fall into one of two
categories: top-down and bottom-up fabrication approaches. In general, the top-down
approach involves semiconductor manufacturing processes and the bottom-up
approach requires chemical and/or biological techniques. The dimensions of struc-
tures fabricated using the top-down approach range from tens of nanometers to a few
hundred microns whereas bottom-up approaches can produce structures down to a
few angstroms in size. These characteristics make top-down approaches generally
more suitable for device platform and bottom-up approaches suitable for sensing
material development. For example, a micromachined diaphragm using a top-down
approach may be used as a device platform and a gas-specific film may be coated
using a bottom-up approach for sensing material. With the adsorption of the gas
onto the sensing film, the resonant frequency of the diaphragm will be altered due
to its mass or stress change.
W.-C. Tian
GE Global Research Center, 1 Research Circle, KW C1324, Niskayuna, NY12309
e-mail: tianw@research.ge.com
66 W.-C. Tian
In this chapter, we first introduce various micro- and nanomachining technolo-

gies for transducer fabrication using different types of materials. The integration of
the sensing material with the transducer to form a sensing element is presented
subsequently.
Micro- and Nanostructuring
Lithography and pattern transfer are two major steps in the fabrication process of a
micromachined transducer. In this section, numerous lithographic techniques,
Section “Lithography,” and pattern transfer approaches, Section “Pattern Transfer,”
for various materials and applications are discussed and compared
Lithography
Lithography is a technique to transfer patterns from either a master hard mask or a

master soft mask (computer layout) to a photon or charged-particle sensitive polymer
film (known as resist) coated on the substrate to which the pattern ought to be trans-
ferred. Other lithographic approaches in addition to optical or charged-particle-based
lithographic techniques are also discussed. Different technologies for optical, charged-
particle, and other lithographic approaches are summarized in Table 4.1.
Optical Approaches
Photolithography is the most commonly used technique/method for pattern transfer.

Typically a quartz plate with Cr pattern is used as the master mask. A photosensitive
resist is spun onto the substrate, to which the pattern needs to be transferred, and is
thermally baked at elevated temperature before exposure to light at the appropriate
wavelength. Light at short wavelengths is typically used for photolithography to
enable high-resolution pattern transfer and avoid running into diffraction limitations
with small features. The quartz mask and the photoresist-coated substrate are then
loaded into the aligner to perform alignment and ultraviolet (UV) exposure.
Depending on the optics used for light exposure, three types of optical lithography
methods can be distinguished. These methods are contact printing, proximity print-
ing, and projection printing. Following the UV exposure step, an optional postbake
can be performed followed by the photoresist development. During the photoresist
development, the exposed areas are etched away, if the resistive is positive, or main-
tained if the photoresist is negative. The light exposure dosage, including light
intensity and exposure time, and the development conditions, including the purity of
developer and development time, are key parameters that affect the photoresist vertical
wall profile and hence the quality of pattern transfer.
4 Integrated Micromachining Technologies for Transducer Fabrication 67
Table 4.1 Summary of Various Lithography Technologies for Micro- and Nanopatterning
Lithography Type Technologies
Optical approaches Ultraviolet (Haefliger and Boisen 2005), deep ultraviolet (Rothschild
2005; Kim and Kim 2005; Liu et al. 2005), extreme ultraviolet
(Collins and Edwards 2006; Gallatin et al. 2003; Silverman 2005;
Tanaka et al. 2005), interference lithography (Chan et al. 2005;
Gutierrez-Rivera et al. 2006; Sewell et al. 2005; Solak 2006),
immersion lithography (Dario et al. 2005; Smith et al. 2005),
zone-plate-array lithography (McGeoch 2005; Menon et al. 2005b)
Charged-particle- Electron beam lithography (Ahmed 1986; Baek et al. 2005; Classen
beam approaches et al. 1992; Kim HS et al. 2006; Myers et al. 2006; Zlobin 2005;
Weber et al. 2005), electron projection lithography (Dhaliwal et al.
2001; Doering et al. 2005; Eguchi et al. 2005; Kasahara et al. 2005;
Koba et al. 2005), ion beam lithography (Gamo and Namba 1984)
Other lithographic X-ray (Di Fabrizio et al. 2004; Jian et al. 2003; Liu et al. 2006; Ueno
approaches et al. 1999), LIGA (Ehrfeld et al. 1988; Guckel 1998; Rogner et al.
1992) (X-ray and LIGA can also be considered optical), scanning
probe lithography (Quate 1997; Soltys et al. 2003) (scanning probe
is also electron-based), hybrid lithography (Cleaver and Ahmed
1984; Steen et al. 2005, 2006; Yang 2005; Yasin et al. 2005)
Mercury arc lamps and excimer lasers are conventional light sources for litho-
graphic tools and emission lines at either 405 or 365 nm wavelengths are commonly
used. At these wavelengths, the minimum resolved line width can vary from 1 um
to a few micrometers (Haefliger and Boisen 2005). To further increase line width
resolution, advanced lithographic tools using deep ultraviolet (DUV) light and
extreme ultraviolet lithography (EUV) with wavelengths shorter than 248 nm are
required. Submicron line widths down to tens of nanometers are achievable with
these techniques (Rothschild 2005; Collins and Edwards 2006; Gallatin et al. 2003;
Silverman 2005; Tanaka et al. 2005; Salib et al. 2005; Liu et al. 2005). EUV lithog-
raphy systems use light with 10–14 nm wavelengths producing feature sizes as
small as 12 nm (Naulleau et al. 2005). Because the wavelengths used for EUV
lithography approach X-ray wavelengths, it is also called soft X-ray lithography.
Photoresist patterning using interference lithography has also been demonstrated
(Chan et al. 2005; Gutierrez-Rivera et al. 2006; Sewell et al. 2005; Solak 2006). An
interference pattern between two or more coherent light sources is formed and
applied to the photoresist layer coating the substrate. Interference lithography
offers advantages over other lithography techniques because of its ability to define
grid patterns over large areas in a single, fast, mask-less exposure.
Immersion lithography is used to extend the resolution of optical lithography
below 100 nm by replacing an air gap between the optics and the wafer with a
liquid (Dario et al. 2005; Smith et al. 2005). This results in dividing the diffraction-
limited feature size by the refractive index of the liquid which is chosen to be
greater than one. The shorter the wavelength of the light used in immersion lithog-
raphy the smaller is the minimum achievable feature size.
68 W.-C. Tian
Maskless optical lithography using MEMS-based spatial-light modulators

(McGeoch 2005; Menon et al. 2005) is an alternative approach to conventional
lithography. This technique, also known as zone-plate-array-lithography (ZPAL),
uses an array of high-numerical-aperture diffractive lenses to create an array of
tightly focused spots on a photosensitive surface. The light intensity for each zone
plate is modulated by a spatial-light modulator. With the latest technology, the
ZPAL technology can match feature resolution that regular mask-based scanners
can produce (65 nm; Martinsson et al. 2005).
Charged-Particle-Beam Approaches
Conventional charged-particle-beam lithographic techniques include electron-beam

lithography (EBL; Ahmed 1986; Baek et al. 2005; Classen et al. 1992; Kim HS
et al. 2006; Myers et al. 2006; Zlobin 2005; Weber et al. 2005) and ion-beam litho-
graphy (IBL; Gamo and Namba 1984). Instead of using a light beam to expose a
photosensitive resist film and transfer the pattern, charged particles (ions or elec-
trons) are used in conjunction with a charged-particle sensitive resist. These tech-
niques are capable of producing high-resolution micro and nano features. EBL and
IBL offer the advantages of maskless exposure, submicrometer resolution, and
precise dimensional control. The overall processing steps of charged-particle-beam
lithography are very similar to steps followed in photolithography.
In order to overcome the diffraction limitation of light in photolithography,
electron-beam lithography provides a high-resolution pattern transfer technique
using high-energy electron beams (100 eV to 100 KeV) to expose electron-sensitive
resist, such as polymethyl methacrylate (PMMA). The EBL resolution is not only
affected by the focused beam spot size but also highly influenced by the scattering
inside the resist and backscattering from the substrate. To compensate for the scat-
tering and backscattering problems, proximity correction algorithms provided with
EBL systems are typically utilized. One disadvantage EBL suffers is the low
throughput. The direct writing time for large-area patterning is long and is not
acceptable for some applications. Electron-beam projection lithography (EBPL)
has been developing to overcome this limitation (Dhaliwal et al. 2001; Doering
et al. 2005; Eguchi et al. 2005; Kasahara et al. 2005; Koba et al. 2005). Few
approaches, such as using projection reduction exposure with variable axis immer-
sion lenses (Dhaliwal et al. 2001), or the large window-size membrane (Eguchi
et al. 2005) have been demonstrated to increase the throughput.
In contrast to EBL, IBL uses a high-energy ion beam instead of an electron beam
to expose the resist. Compared to any other lithographic approaches, IBL provides the
smallest beam spot size, down to 8 nm, and the generated ion energy is lower with
negligible scattering in resist and backscattering from the substrate. These features
enable the best resolution among all lithographic techniques (Gamo and Namba
1984). However, as with EBL, the long direct writing time and high vacuum exposure
requirement became the major limitation for this charged-particle-beam technique.
Other Lithographic Approaches
X-ray lithography provides several advantages over optical lithography. Because of

the short wavelengths (a few angstroms) generated by a synchrotron radiation
source, the X-ray lithography system offers very small feature size capability, high
reproducibility independent of substrate material, controllable surface topology,
and immunity from small particle effects (Jian et al. 2003; Ueno et al. 1999; Di
Fabrizio et al. 2004; Liu et al. 2006).
In addition to X-ray lithography, a technique called LIGA, the German acronym
for X-ray lithography (X-ray LIthographie), electrodeposition (Galvanoformung),
and molding (Abformtechnik), has been implemented for pattern transfers. LIGA is
also known as deep X-ray lithography and is particularly useful to fabricate high
aspect ratio microstructures with a thickness in the range of few micrometers to
several centimeters. Various microelectromechanical systems (MEMS) have been
fabricated using this technique (Ehrfeld et al. 1988; Guckel 1998; Rogner et al.
1992), as shown in Fig. 4.1.
Scanning probe lithography uses scanning tunneling microscopes or atomic force
microscopes to pattern nanometer-scale features (Quate 1997; Soltys et al. 2003).
Fig. 4.1 Thick microstructures fabricated using LIGA (Guckel 1998). (Reproduced by permis-
sion from IEEE, copyright 1998, IEEE.)
70 W.-C. Tian
When a voltage bias is applied between a sharp probe tip and the substrate, an
electric field is generated around the tip and can be used to locally oxidize Si or to
expose resist (Quate 1997). This local oxidation process is powerful because of its
fine resolution (sub-50 nm) and the robust oxide etch mask that is created. In addi-
tion, the electrons generated around the tips have lower energy (<50 eV) compared
to EBL (100 eV to 100 KeV). This lower energy prevents the scattering in the resist
and backscattering from the substrate and thus enhances the resolution.
Hybrid lithography, either by combining electron/ion beam lithography or opti-
cal/electron beam lithography, has been successfully implemented (Cleaver and
Ahmed 1984; Steen et al. 2005, 2006; Yang ; Yasin et al. 2005). These techniques
use different lithographic methods, for example, optical lithography and EBL, on a
single resist and marry the advantages of optical lithography (high throughput) and
EBL (high resolution) (Steen et al. 2006).
Pattern Transfer
Lithographic methods transfer patterns either from a hard or a soft mask (computer
layout charged-particle-beam lithography) to the resist coating the substrate, which
can be semiconductor and polymeric materials. The resist typically provides an
etch mask for further subtractive (etching) or additive (deposition) processes used
to build the transducer. In this section, various micro- and nanopattern transfer
technologies for numerous microsensor platform materials are introduced.
Semiconductor Material Micro/Nanomachining
Semiconductor (Si, SiC, GaN, etc.) and ceramic materials (e.g., glass) are common
substrates for micromachined transducers (Peterson 1982). Fabrication techniques
that involve etching of bulk wafers are known as bulk micromachining whereas
fabrication techniques that involve deposition of thin and thick films of materials
(e.g., polycrystalline silicon, silicon nitride) are known as surface micromachining.
Several pattern transfer methods, such as plasma etching, wet etching, laser abla-
tion, ion milling, and sand blasting for both bulk and surface micromachining are
discussed in this section.
Plasma etching is one of the most effective methods for pattern transfer and is
independent of crystal orientation. A photoresist, oxide or metal layer, is typically
used as the etch mask depending on the chemistries of the used etching technolo-
gies. Fluorine-based plasma etching with a resist mask can be employed to fabricate
high aspect ratio (thick) structures whereas chlorine-based plasma etching with a
metal etch mask can be implemented for creating submicrometer (thin) structures
(Rakhshandehroo et al. 1998; Tian and Pang 2001, 2002, 2003, Tian et al. 2000).
Many plasma etching processes using fluorine chemistry have been developed
to produce various microstructures. The advantages of fluorine etching include high
etch rates and high selectivity to etch masks. However, fluor atoms react with
silicon spontaneously and result in isotropic silicon etching to cause undercutting
underneath the mask. A process, known as the Bosch process, utilizes a multiplexed
etch (sulphur hexafluoride, SF6)/passivation (Octafluorocyclobutane, C4F8) cycle to
alleviate undercutting issues (Robert Bosch GmbH 1990 and 1991). Wafer cooling
is another solution to reduce undercutting problems associated with fluorine-based
plasma etching (Tachi et al. 1991).
Chlorine-based plasma etching is ion-assisted rather than spontaneous and is
therefore anisotropic (Rakhshandehroo et al. 1998). The sidewalls of microstruc-
tures fabricated by chlorine-based etching are very vertical and no sidewall passiva-
tion or wafer cooling is required compared to fluorine-based etching. The disadvantage
of chlorine-based etching is a lower etch rate and selectivity to the etch mask in
comparison to fluorine-based etching.
Wet chemical etchants, such as potassium hydroxide (KOH), tetramethylammo-
nium hydroxide (TMAH), or ethylene diamine pyrocatechol (EDP), in conjunction
with a heavily doped layer used as an etch stop in semiconductor materials can also
be used to form microstructures (Kovacs et al. 1998). A p++ doped Si layer with a
thickness ranging from 3–15 um can be utilized to define the thickness of the
microstructures because the etchants will not attack heavily doped semiconductors.
Other etch stop examples include silicon nitride or a multiple-layer film combining
silicon oxide and silicon nitride. A plasma etch step can be combined with various
etch stop technologies to fabricate freestanding microstructures. Although struc-
tures fabricated by wet etching generally provide a better surface roughness and
lower cost compared to plasma etching technologies, the crystal orientation
dependence limits wet etching abilities to produce complex geometries.
Glass is also a very common choice for device substrates. Wet etching processes
using chemicals such as hydrofluoric acid (HF) or buffer hydrofluoric acid (BHF),
have been applied to micromachine glass substrates (Corman et al. 1998). However,
these etching methods are isotropic and the etch mask cannot resist long for build-
ing deep trenches or channels. Dry etching techniques, such as laser ablation
(Klank et al. 2002, Liu et al. 1997, Schaffer et al. 2001), ion milling, or sand blast-
ing (Steve 2001, Wensink et al. 2000) can be used to pattern structures in glass and
overcome wet etching limitations.
Ablation is defined as the removal of material from the surface of a substrate by
vaporization, chipping, or other erosive processes. Various laser systems (carbon
dioxide, CO2 or neodymium-yttrium aluminium garnet, Nd-YAG, laser) have been
successfully employed for micro-/nanomachining (Klank et al. 2002; Liu et al.
1997). Very high aspect ratios can be achieved by this technique but the associated
capital investment may be too high.
Ion milling entails high-energy inert ion bombardment on the substrate to physi-
cally break the bonding between molecules at the surface of the substrate and then
remove the substrate molecules by momentum transfer (Reyntjens et al. 2001). The
ion energy needed to break the molecular bonding of the substrate material is typi-
cally much larger compared to the chemical bond energy of substrate materials, so
the material dependence etch rate is minimal. One disadvantage of ion milling is
72 W.-C. Tian
that achievable etch rates are typically lower compared to other etching
techniques.
Sandblasting is a technology where solid particles are accelerated in a focused
beam and used to hit the hard surface of a substrate at very high speed to smooth,
clean, or shape the surface of the substrate. Recently, it has been introduced for
micro-/nanomachining and has been used to demonstrate through wafer vias with
high aspect ratio(Schlautmann 2001; Wensink et al. 2000). Although the cost of this
technology is low, the ability to pattern very fine features is limited to the microm-
eter range and the sidewall profile is typically tapered. These three dry etching
technologies can be used not only for glass, but also for other semiconductor mate-
rials such as silicon (Si).
Micro-/Nanomachining of Polymeric and Plastic Materials
Polymer processes are in general lower cost compared to semiconductor-type

processes. Sensing devices that are geared towards life sciences applications
tend to use polymers because they enable disposable biomedical markets, such
as blood testing or biological sensing and detection, and the like (Chong et al.
2004). In this section, several polymer/plastic patterning technologies, such as
nanoimprinting, hot embossing, and injection molding, are discussed and
compared.
Thermopress imprinting and “step and flash” imprinting are two very promising
technologies in this field. Thermopress nanoimprint lithography was first devel-
oped and demonstrated by Stephen Y. Chou’s group (Chou 2001). A thin layer of
resist, thermal plastic material such as polymethyl methacrylate (PMMA), is spun
on a substrate and a mold is used to depress the resist at elevated temperatures to
form micro-/nanostructures on the substrate. Compared to conventional techniques,
such as EBL, nanoimprinting provides a much faster process to pattern nanostruc-
tures over large areas.
Step and flash imprint lithography (SFIL) using photopolymerization was devel-
oped by Grant Willson’s group (Resnick et al. 2005). In this technique, UV-curable
resist is applied to the substrate and the mold is used to depress the resist followed
by UV light exposure to harden the resist. The mold is normally made of transpar-
ent materials such as fused silica or glass.
For the last two techniques, an optional anisotropic etch step can be utilized to
remove the residue polymer from the substrate surface. Therefore, the polymer pat-
terns now become the etch mask and the patterns can be transferred to the underly-
ing substrate. Currently, sub-10 nm features in PMMA on Si or a metal substrate
with excellent uniformity have been demonstrated (Chou 2001). Fig. 4.2 shows a
nanodot array formed by imprinting into a PMMA resist followed by a lift-off
process.
Hot embossing is the stamping of patterns from a micro-/nanomachined mold
into a polymer softened above its glass transition temperature (Juang et al. 2002;
Holger and Ulf 2000; Scheer et al. 2001; Gerlach et al. 2002). The molds are
fabricated by the micro-/nanomachining technologies mentioned in previous sections.
Fig. 4.2 SEM photos of a top view of 10 nm diameter metal dots (Chou 2001). (Reproduced by
permission from MRS Bulletin.)
A wide variety of polymer materials, such as cycloolefin-copolymer (COC), poly-

carbonate (PC), polyethylene (PE), polyetherether-ketone (PEEK), PMMA, poly-
styrene (PS), and polyvinyl butyral (PVB), can be used for hot embossing to
fabricate micro-/nanostructures.
Injection molding techniques, using similar principles as hot embossing, provide
a high-throughput approach by automation of the whole stamping process.
Originally, this technology was developed for shaping macroplastic devices but has
been recently adopted for microdevice fabrication in various areas, such as micro-
fluidics (Becker and Gartner 2000). The automation of injection molding makes
micro-/nanofabrication very low cost. Examples of such use include the fabrication
of compact disks (CD) or lab-on-a-chip (LOC; Lee et al. 2001). The mold and poly-
mer materials used for injection molding are also very similar to those used in hot
embossing methods.
Micro- and Nanostructure and Device Integration
Various micro-/nanomachining technologies mentioned in previous sections can be

applied to fabricate numerous micro-/nanostructures used to build transducers. The
integration of the sensing film/nanostructures with a micromachined transducer can
74 W.-C. Tian
occur simultaneously or subsequent to the fabrication of the transducer. In this sec-

tion, various sensor platforms and integration techniques are discussed assuming
the sensing structures are added to a developed transducer. The sensor operation
principle is beyond the scope of this section and the readers can refer to other
chapters.
Micro- and Nanostructures for Sensing
Numerous measurands can be targeted with micromachined sensors. Depending on

the application and the measurand, different transducers can be needed including
acoustic, optical, electrical, magnetic, mechanical, or combinations of the above.
Three types of transducers are discussed in this section: electrode-based, semicon-
ductor-based, and MEMS-based, as summarized in Table 4.2.
Electrode-Based Platform
Electrode-based transducers are the most popular because of their simple fabrication
processes and ease of sensing film integration. Metal thin films are typically depos-
ited and patterned on top of a substrate to form electrodes in various patterns. The
sensing material (thin film or nanostructures) is then locally incorporated in between
electrodes or on top of the electrodes and the sensor output is measured in the form
of impedance (Han et al. 2005; Lapicki et al. 2005; Park et al. 2005; Sun et al. 2006),
voltage (Lapicki et al. 2005; Park et al. 2005), conductance (Hernandez et al. 2005;
Sayago et al. 2005; Smardzewski et al. 2004), or frequency (Ippolito 2005) change.
Examples of electrode designs include interdigitated electrodes (Han et al. 2005;
Smardzewski et al. 2004; Ippolito 2005) as well as simple square contact pads
(Hernandez et al. 2005). The substrate material selection is often dictated by the
application. A glass or quartz substrate can be used for chemiresistive measurement
(Han et al. 2005; Lapicki et al. 2005; Park et al. 2005; Sun et al. 2006) and a gallium
arsenide (GaAs) substrate can be used for Hall effect sensing (Lapicki et al. 2005).
Table 4.2 Summary of Various Transducers

Transducer Sensor Output
Electrode-based Resistance (Sun et al. 2006; Park et al. 2005; Lapicki et al. 2005;
Han et al. 2005), voltage (Lapicki et al. 2005; Park et al. 2005),
conductance (Hernandez et al. 2005; Sayago et al. 2005 and
frequency (Ippolito 2005)
Semiconductor-based Trans-conductance (Bangar et al. 2005; Modi and Lacy 2005),
surface stress (Braun et al. 2006), frequency (Kim et al. 2004;
Zribi et al. 2005)
MEMS-based Voltage (Modi and Lacy 2005), surface stress (Braun et al. 2006),
frequency (Kim et al. 2004; Kim DS et al. 2006; Zribi et al. 2005)
Other substrates, such as alumina for conductance measurement and lithium nio-
bate (LiNbO3) for surface acoustic wave sensing, have also been reported recently
(Ippolito 2005; Sayago et al. 2005).
Semiconductor-Based Platform
The most common semiconductor-based transducer is the field effect transistor

(FET). There are three electrodes: drain, source, and gate, in a FET-based transducer
(Kim DS et al. 2006). In the case of a silicon-based FET, the drain and source elec-
trodes are deposited and patterned on top of oxidized Si and a conductive (metal or
heavily doped Si) thin film serves as the gate electrode. The sensing material is typi-
cally placed on top of the gate electrode or the gate electrode can be extended to a
microfluidic channel for liquid-based sensing, as shown in Fig. 4.3. The substrate
material for this type of transducer is typically a semiconductor.
At a given gate voltage, the current and voltage between source and drain can be meas-
ured and the trans-conductance can be calculated. The trans-conductance will change
with ambient environment change, for example, humidity or gas (Bangar et al. 2005).
MEMS-Based Platform
MEMS-based transducers involve mechanical, electrical, and electromagnetic

micro- or nanostructures with the sensing materials. The transducer response to
analytes can be a resonant frequency change, impedance change, or other parame-
ters (Braun et al. 2006; Modi and Lacy 2005; Zribi et al. 2005).
Fig. 4.3 Top view of a FET-based transducer (Kim DS et al. 2006). (Reproduced by permission
from Elsevier, copyright 2006, Elsevier.)
76 W.-C. Tian
For a resonant MEMS transducer, the natural frequency of the device is a

function of the structural materials properties and various structure geometries.
Some typical microstructures can be cantilever beams, clamp–clamp beams or
bridges, and diaphragms, and the typical materials used are Si, poly-Si, and sili-
con nitride.
For a MEMS-based impedance sensor platform, a gap in between an electrical
conductive cantilever beam or a diaphragm and an electrode on the substrate will
be formed. These transducers are fabricated through conventional micromachining
technologies, such as lithography, etching, and sensing film deposition as we dis-
cussed in previous chapters.
Sensitive Material Integration
Nanostructure assembly (immersion coating assembly, layer-by-layer coating,

other self-assembly techniques), deposition, localized polymerization, and spotting
are the most commonly used techniques to integrate sensing materials with
micromachined transducers, as summarized in Table 4.3.
Nanostructure Assembly
Immersion coating assembly is used to integrate functional thin films and nanostruc-
tures with interdigitated microelectrode (IME) transducers. The IME transducer is
fabricated with metallization followed by lithography. This IME substrate is subse-
quently immersed into the prepared coating sol or the solution of mixed nanoparti-
cles and thiols (Han et al. 2005; Lapicki et al. 2005; Park et al. 2005; Sun et al.
2006). The thickness of the functional material can be controlled by the withdrawing
rate, immersion time, or the drying temperature. The multilayer can be obtained by
several repetitions of coating steps. Alcohol gas sensing using IME transducers with
titanium oxide (TiO2) film and detection of volatile organic compounds or nitroaro-
matic compounds using nanoparticles/thiols on IME transducers have been demon-
strated (Han et al. 2005; Lapicki et al. 2005; Park et al. 2005; Sun et al. 2006).
Layer-by-layer (LBL) assembly is also a common technique for functional materials
integration with the transducer. This assembly technique provides a simple, versatile,
and inexpensive approach for thin-film formation by alternating the deposition of
oppositely charged species from an aqueous solution (Hammond 2004). LBL has
been successfully used to integrate sensing materials such as polyelectrolytes and
TiO2 nanoparticle films for gas and humidity sensing (Kim et al. 2004), polyacrylic
acid/polyethylene oxide film for pH sensing (Lutkenhaus et al. 2005), and den-
drimer/metallic nanoparticles film for biosensing (Goulet et al. 2005). A combina-
tion of lithographic approach and LBL assembly is demonstrated to spatially
pattern the polystyrene particles (150 and 64 nm) with feature sizes of 5 to 20 um
(Cui et al. 2004).
Table 4.3 Summary of Integration Methods of Sensing Materials and Micromachined Transducers
Integration Technology Functional Material
Nanostructure assembly Immersion coating assembly
Immersion coating assembly (Sun et al. 2006) Titanium oxide (TiO2)
Immersion (Han et al. 2005) Nanoparticles and thiols
Layer-by-layer (LBL) assembly
LBL (Hammond 2004) Various materials
LBL (Kim et al. 2004) Polyelectrolytes and TiO2 nanoparticles
LBL (Lutkenhaus et al. 2005) Polyacrylic acid (PAA); polyethylene
oxide (PEO); dye
LBL (Goulet et al. 2005) Dendrimer/metallic nanoparticles on glass
LBL and photolithography (Cui et al. 2004) Polystyrene nanoparticles
Other self-assembly
Magnetically aligned (Hernandez et al. 2005) Iron palladium (FePd) alloy nanowires
Biological-based self-assembly Zirconium dioxide (ZrO2) nanoparticles
(Zhao G et al. 2005) with heme proteins on functional
glassy carbon electrode
Nanostructured deposition
Electrochemical Deposition (Park et al. 2005) Nanoporous platinum oxide (PtO)
Deposition metal–insulator–metal ensemble Alkanethiol–Au nano cluster
(Smardzewski 2004)
Sputtering (Ippolito 2005) Zinc oxide/indium oxide (ZnO/InOx) on
XZ lithium niobate (LiNbO3) substrate
Sputtering (Zhao Z et al. 2005) 60% Palladium (Pd) 40% gold (Au)
Sputtering (Mardare et al. 2005) Chrome (Cr)-doped titanium oxide
Plasma enhanced chemical vapor deposition Metal-containing diamondlike
(PECVD) for carbon-silicon and sputtering nanocomposite
for metal (Takeno et al. 2005)
Pulsed laser deposition (Modi and Lacy 2005) ZnO nanorods
Pulsed laser deposition (Vainos 2004) Purely epitaxial nanocomposite
Glancing angle deposition (Kim DS et al. 2006) Silicon dioxide
Thermal evaporation of metallic tungsten oxide Tungsten oxide (WO3)
powder followed by thermal annealing
(Jayatissa and Gupta 2005)
MBE and lithography (Lapicki et al. 2005) Indium antimonide (InSb)
Polymerization
Electropolymerization (Bangar et al. 2005; Fan Monomer nanowires
and Lu 2005; Hammond 2004)
Spin coating and UV polymerization (Schmidt Molecularly imprinted polymer
and Haupt 2005)
Spotting
Carbon nanotubes (CNTs) were dissolved in ethanol Single-walled carbon nanotube (SWCNTs)
and sprayed by air brush (Sayago et al. 2005) with palladium doping
Ink jet spotting (Braun et al. 2006) Proteoliposomes
Pipetting (Zribi et al. 2005) Polystyrene sulfonic acid and carbon
nanotubes
78 W.-C. Tian
Other self-assembly techniques, such as magnetic-based or biological, have

been used for various applications. Iron palladium (FePd) alloy nanowires are mag-
netically aligned to microferromagnetic electrodes for hydrogen sensing (Hernandez
et al. 2005). Self-assembly of zirconium dioxide (ZrO2) nanoparticles with heme
proteins on functional glassy carbon electrodes were developed for hydrogen per-
oxide sensing (Zhao G et al. 2005).
Nanostructure Deposition
Nanostructures can be formed using various processes, such as electrochemi-

cal deposition (Park et al. 2005), sputtering (Ippolito 2005; Mardare et al.
2005; Zhao Z et al. 2005), chemical vapor deposition (Fan et al. 2005;
Smardzewski et al. 2004; Takeno et al. 2005), pulsed laser deposition (Modi
and Lacy 2005; Vainos 2004), glancing angle deposition (Kim DS et al. 2006),
thermal evaporation (Jayatissa and Gupta 2005), and molecular beam epitaxy
(MBE) (Lapicki et al. 2005).
For electrochemical deposition, nanoporous platinum oxide films have been
demonstrated for pH sensing (Park et al. 2005). The use of sputtering techniques
has also been reported in several applications. Metal oxides, such as zinc oxide/
indium oxide (ZnO/InO) have been deposited by radio frequency (RF) magnetron
sputtering/direct current (dc) sputtering on LiNbO3 substrate for nitrogen oxide and
hydrogen sensing (Ippolito 2005); 60% of Pd and 40% of gold (Au) film has been
deposited by dc magnetron sputtering and have been used for hydrogen sensing
(Zhao Z et al. 2005). Also, undoped and chrome (Cr)-doped titanium oxide films
have been deposited by dc reactive sputtering for gas-sensing applications (Mardare
et al. 2005).
Pulsed laser deposition techniques have been used to prepare functional thin
films, such as ZnO nanorods for gas/chemical sensing (Modi and Lacy 2005)
and epitaxial nanocomposites for physicochemical sensing (Vainos 2004). For
the thermal evaporation method, evaporation of metallic tungsten oxide powder
followed by thermal annealing has been proved for gas sensors (Jayatissa and
Gupta 2005).
The indium antimonide (InSb) thin film for the Hall sensor is grown on a GaAs
substrate using the MBE approach (Lapicki et al. 2005). A combination of lithogra-
phy, etching, and metal contact deposition is used to fabricate this sensor platform.
Polymerization
Polypyrrole films doped with poly-aminobenzene sulfonic acid (PABS) functional-

ized single-walled nanotubes (SWNT) (PPy/SWNT-PABS) have been reported
recently (Fan and Lu 2005). A 2 mL volume of the electrolyte solution containing
pyrrole and SWNT-PABS in water was introduced between the metal electrodes.
The electropolymerization of the functional thin film took place from the anode to
the cathode under an applied current and can be designed with individual address-
ability. By simply adding biomolecules in the solution, biofictionalization of con-
ducting polymer nanowires can also be implemented (Bangar et al. 2005). Other
polymerization techniques for functional thin film, such as molecularly imprinted
polymer mimicking the behavior of natural antibodies, can be prepared by spin
coating followed by UV polymerization (Schmidt and Haupt 2005).
Spotting
A hydrogen sensor based on carbon nanotubes (CNTs) with Pd doping has been
reported (Sayago et al. 2005). The CNTs were deposited on an alumina substrate
via airbrush. The CNTs were dissolved in ethanol and sprayed on alumina using
dry air carrier gas. The results showed CNTs are p-type semiconductors and the
resistance increased with reducing gas concentration.
A reliable functionalization of biofilm onto a transducer and a successful detec-
tion with quantitative data have been reported by Braun et al. (2006). Application
of proteoliposomes onto MEMS cantilever arrays using ink jet spotting was dem-
onstrated. The authors claimed that this technique can be potentially used to meas-
ure membrane protein-based receptor–ligand interactions and conformational
changes.
Another example using spotting to incorporate functional thin film into a sensor
platform is MEMS resonant gas sensors (Zribi et al. 2005), as shown in Fig. 4.4.
The sensor consists of a resonant nitride membrane and a nanostructured gas-
sensing film with high adsorption and specificity to analytes. The sensing film
solution is deposited via pipetting with a few microliters onto the backside of the
membrane. Thin films of polystyrene sulfonic acid and single-wall carbon nano-
tubes have been demonstrated to detect moisture and CO2 separately using this
transducer.
Fig. 4.4 The photograph on the left shows the top view of the transducer. The drawing on the
right shows the cross-section of the device (Zribi et al. 2005). (Reproduced with permission from
Elsevier, copyright 2005, Elsevier.)
80 W.-C. Tian
Summary
This chapter described various micro- and nanostructuring technologies for

micromachined transducers and integration techniques for the sensing material
onto the transducer. The latest lithography and pattern transfer technologies
for micro- and nanopatterning were introduced. Various methods to integrate
sensitive materials with different micromachined transducers were also
discussed.
References
Ahmed H (1986) Advances in high resolution and high throughput in electron beam lithography.
Vacuum, 37:375.
Baek IB et al. (2005) Electron beam lithography patterning of sub-10 nm line using hydrogen
silsesquioxane for nanoscale device applications. Journal of Vacuum Science and Technology
B: Microelectronics and Nanometer Structures, 23:3120–3123.
Bangar M et al. (2005) Individually addressable conducting polymer nanowires for FET based
label-free sensing, 208th Meeting of The Electrochemical Society – Meeting Abstracts,
200502:2510.
Becker H, Gartner C (2000) Polymer microfabrication methods for microfluidic analytical appli-
cations. Electrophoresis, 21(1):12–26.
Braun T et al. (2006) Conformational change of bacteriorhodopsin quantitatively monitored by
microcantilever sensors. Biophysical Journal, 90:2970–2977.
Chan TYM et al. (2005) Photonic band gap templating using optical interference lithography.
Physical Review E – Statistical, Nonlinear, and Soft Matter Physics, 71:046605.
Chong HA, Choi J-W, Beaucage G, Nevin JH, Lee J-B, Puntambekar A, Le JY (2004) Disposable
smart lab on a chip for point-of-care clinical diagnostics. IEEE Proceedings, 92:154–173.
Chou SY (2001) Nanoimprint lithography and lithographically induced self-assembly. MRS
Bulletin, 26:512.
Classen A et al. (1992) High voltage electron beam lithography of the resolution limits of SAL
601 negative resist. Microelectronic Engineering, 17:21–24.
Cleaver JRA, Ahmed H (1984) Combined electron and ion beam lithography system. Journal of
Vacuum Science & Technology B: Microelectronics Processing and Phenomena,
3:144–147.
Collins L, Edwards C (2006) Tricks of the light. IEE Electronics Systems and Software,
4:42–45.
Corman T et al. (1998) Deep wet etching of borosilicate glass using an anodically bonded silicon
substrate as mask. Journal of Micromechanics and Microengineering, 8:8–87.
Cui T et al. (2004) Lithographic approach to pattern multiple nanoparticle thin films prepared by
layer-by-layer self-assembly for microsystems. Sensors and Actuators, A: Physical,
114:501–504.
Dario G et al. (2005) First microprocessors printed with immersion lithography. Microlithography
World, 14:4, 6, 18.
Dhaliwal RS et al. (2001) PREVAIL – Electron projection technology approach for next-genera-
tion lithography. IBM Journal of Research and Development, 45:615–636.
Di Fabrizio E et al. (2004) X-ray lithography for micro- and nano-fabrication at ELETTRA for
interdisciplinary applications. Journal of Physics Condensed Matter, 16:S3517–S3535.
Doering H-J et al. (2005) Proof-of-concept tool development for projection mask-less lithography
(PML2). Progress in Biomedical Optics and Imaging – Proceedings of SPIE, Emerging
Lithographic Technologies IX, 5751:355–365.
Eguchi H et al. (2005) Image placement of large window-size membrane for EPL and LEEPL
mask. Proceedings of SPIE – The International Society for Optical Engineering, Photomask
and Next-Generation Lithography Mask Technology XII, 5853:910–920.
Ehrfeld W et al. (1988) LIGA process: Sensor construction techniques via x-ray lithography.
Technical Digest, IEEE Solid State Sensor and Actuator Workshop, 1–4.
Electrochemical impedance spectroscopy study of mixed thiol monolayers on gold, Xing, Y.F.
(Institute of Material Research and Engineering); Li, S.F.Y.; Lau, A.K.H.; O’Shea, S.J. ,
Journal of Electroanalytical Chemistry, v 583, n 1, Sep 1, 2005, p 124-132
Fan L et al. (2005) The development of functional endohedral metallofullerene materials.
Fullerenes Nanotubes and Carbon Nanostructures, 13:155–158.
Fan ZY, Lu JG (2005) Gate-refreshable nanowire chemical sensors. Applied Physics Letters,
86:123510.
Gallatin GM, Houle FA, Cobb JL (2003) Statistical limitations of printing 50 and 80 nm contact
holes by EUV lithography. Journal of Vacuum Science and Technology B: Microelectronics
and Nanometer Structures, 21:3172–3176.
Gamo K, Namba S (1984) Ion beam lithography. Ultramicroscopy, 15:261–269.
Gerlach AK, Guber G, Heckele AE, Herrmann M, Muslija A, Schaller T (2002) Microfabrication
of single-use plastic microfluidic devices for high-throughput screening and DNA analysis.
Microsystem Technologies, 7:265–268.
Goulet PJG et al. (2005) Surface-enhanced Raman scattering on dendrimer/metallic nanoparticle
layer-by-layer film substrates. Langmuir, 21:5576–5581.
Guckel H (1998) High-aspect-ratio micromachining via deep x-ray lithography. Proceedings of
the IEEE, 86:1586–1593.
Gutierrez-Rivera LE et al. (2006) Nanosieves fabricated by interference lithography and electro-
forming. Proceedings of SPIE – The International Society for Optical Engineering, Device and
Process Technologies for Microelectronics, MEMS, and Photonics IV, 6037:603710.
Haefliger D, Boisen A (2005) Contact printed masks for 3D microfabrication in negative resists.
Proceedings of the IEEE International Conference on Micro Electro Mechanical Systems
(MEMS), Proceedings of the 18th IEEE International Conference on Micro Electro Mechanical
Systems, MEMS 2005 Miami – Technical Digest, 556–559.
Hammond PT (2004) Form and function in multilayer assembly: New applications at the nanos-
cale. Advanced Materials, 16:1271–1293.
Han Li et al. (2005) Nanoparticle-structured sensing array materials and pattern recognition for
VOC detection. Sensors and Actuators, B: Chemical, 106:431–441.
Hernandez S et al. (2005) Electrodeposition of FePd alloy thin films and nanowires. 208th
Meeting of the Electrochemical Society – Meeting Abstracts, MA 2005-02:1221.
Holger B, Ulf H (2000) Hot embossing as a method for the fabrication of polymer high aspect
ratio structures. Sensors and Actuators, 83:130–135.
Ippolito SJ (2005) Highly sensitive layered ZnO/LiNbO3 SAW device with InOx selective layer
for NO2 and H2 gas sensing. Sensors and Actuators, B: Chemical, 111–112:207–212.
Jayatissa AH, Gupta T (2005) Fine-grain nanocrystalline tungsten oxide films for gas sensor
applications. Ceramic Transactions, Ceramic Nanomaterials and Nanotechnology III –
Proceedings of the 106th Annual Meeting of the American Ceramic Society, 159:119–124.
Jian L et al. (2003) SU-8 based deep x-ray lithography/LIGA. Proceedings of SPIE – The
International Society for Optical Engineering, 4979:394–401.
Juang Y-J, Lee LJ, Koelling KW (2002) Hot embossing in microfabrication. Part I: Experimental,
Polymer Engineering & Science, 42(3):539–550.
Kasahara H et al. (2005) Performances by the electron optical system of low energy electron beam
proximity projection lithography tool with a large scanning field. Journal of Vacuum Science
and Technology B: Microelectronics and Nanometer Structures, 23:2754–2757.
82 W.-C. Tian
Kim DS et al. (2006) An extended gate FET-based biosensor integrated with a Si microfluidic chan-
nel for detection of protein complexes. Sensors and Actuators B: Chemical, 117:488–494.
Kim HS et al. (2006) Low energy electron beam microcolumn lithography. Microelectronic
Engineering, 83:962–967.
Kim JB, Kim K-S (2005) Novel photobleachable deep UV resists based on single component non-
chemically amplified resist system. Macromolecular Rapid Communications, 26:1412–1417.
Kim JH et al. (2004) Fabrication of nanoporous and hetero structure thin film via a layer-by-layer
self assembly method for a gas sensor. Sensors and Actuators, B: Chemical, 102:241–247.
Klank H, Kutter JP, Geschke O (2002) CO2-laser micromachining and back-end processing for
rapid production of PMMA-based microfluidic systems. Lab on a Chip, 2:242–246.
Koba F et al. (2005) EPL performance in 65-nm node metallization technology and beyond.
Progress in Biomedical Optics and Imaging – Proceedings of SPIE, Emerging Lithographic
Technologies IX, 5751:501–508.
Kovacs GTA, Maluf NI, Peterson KE (1998) Bulk micromachining of Si. Proceedings of IEEE,
86:1536–1551.
Lapicki A et al. (2005) Functionalization of micro-hall effect sensors for biomedical applications
utilizing superparamagnetic beads. IEEE Transactions on Magnetics, 41:4134–4136.
Lee LJ, Madou MJ, Koelling K, Daunert W, Lai S, Koh S, Juang CG, Lu Y-J, Yu Y (2001) Design
and fabrication of CD-like microfluidic platforms for diagnostics: Polymer-based microfabri-
cation. Biomedical Microdevices, 3:339–351.
Liu G et al. (2006) Fabrication of 3D photonic crystal by deep x-ray lithography. Proceedings of
SPIE – The International Society for Optical Engineering, 6110:61100R.
Liu X, Brenner K-H, Wilzbach M, Schwarz M, Fernholz T, Schmiedmayer J (2005) Fabrication
of alignment structures for a fiber resonator by use of deep-ultraviolet lithography. Applied
Optics, 44:6857–6860.
Liu X, Du D, Mourou G (1997) Laser ablation and micromachining with ultrashort laser pulses.
IEEE Journal of Quantum Electronics, 33:1706–1716.
Lutkenhaus JL et al. (2005) Elastomeric flexible free-standing hydrogen-bonded nanoscale assem-
blies. Journal of the American Chemical Society, 127:17228–17234.
Mardare D et al. (2005) Chromium-doped titanium oxide thin films. Materials Science and
Engineering B: Solid-State Materials for Advanced Technology, Symposium D: Functional
Oxides for Advanced Semiconductor Technologies, 118:187–191.
Martinsson H et al. (2005) Current status of optical maskless lithography. Journal of
Microlithography, Microfabrication and Microsystems, 4:1–15.
McGeoch MW (2005) Maskless EUV lithography via optically addressed modulator. Progress in
Biomedical Optics and Imaging – Proceedings of SPIE, Emerging Lithographic Technologies
IX, 5751:349–354.
Menon R et al. (2005a) Maskless optical lithography using MEMs-based spatial-light modulators.
Progress in Biomedical Optics and Imaging – Proceedings of SPIE, MOEMS Display and
Imaging Systems III, 5721:53–63.
Menon R et al. (2005b) Maskless lithography. Materials Today, 8:26–33.
Metal-insulator-metal ensemble (MIME) chemical detectors, Smardzewski, R.R. (Geo-Centers,
Inc., Gunpowder Branch); Jarvis, N.L.; Snow, A.W.; Wohltjen, H.; Mackay, R.A. , 2004 NSTI
Nanotechnology Conference and Trade Show - NSTI Nanotech 2004, v 3, 2004 NSTI
Nanotechnology Conference and Trade Show - NSTI Nanotech 2004, 2004, p 163-164
Modi N, Lacy F (2005) Piezoelectric microcantilevers of nanoscale thickness for detection of
cells. 2005 NSTI Nanotechnology Conference and Trade Show – NSTI Nanotech 2005
Technical Proceedings, 2005 NSTI Nanotechnology Conference and Trade Show – NSTI
Nanotech 2005 Technical Proceedings, 1:501–504.
Myers BD et al. (2006) Variable pressure electron beam lithography (VP-eBL): A new tool for
direct patterning of nanometer-scale features on substrates with low electrical conductivity.
Nano Letters, 6:963–968.
Naulleau P et al. (2005) EUV microexposures at the ALS using the 0.3-NA MET projection
optics. Progress in Biomedical Optics and Imaging – Proceedings of SPIE, Emerging
Lithographic Technologies IX, 5751:56–63.
Park S et al. (2005) pH-sensitive solid-state electrode based on electrodeposited nanoporous plati-
num. Analytical Chemistry, 77:7695–7701.
Peterson KE (1982) Silicon as a mechanical material. Proceedings of IEEE, 70:420–457.
PZT stack etch for MEMS devices in a capacitively coupled high density plasma reactor”,
Werbaneth, Paul (Tegal Corporation); Almerico, John; Jerde, Les; Marks, Steve, Proceedings
of SPIE - The International Society for Optical Engineering, v 4592, 2001, p 267-273
Quate CF (1997) Scanning probes as a lithography tool for nanostructures. Surface Science,
386:259–264.
Rakhshandehroo MR, Weigold JW, Tian WC, Pang SW (1998) Dry etching of Si field emitters
and high aspect ratio resonators using an inductively coupled plasma source. Journal of
Vacuum Science and Technology B, 16:2849–2854.
Resnick DJ, Sreenivasan SV, Willson CG (2005) Step and flash imprint lithography. Materials
Today, 8(2):34–42.
Reyntjens S et al. (2001) A review of focused ion beam applications in microsystem technology.
Journal of Micromechanics and Microengineering, 11:287–300.
Robert Bosch GmbH (1990, 1991) Trench etch process for a single wafer RIE dry etch reactor,
U.S. Patent 4,855,017 and 4,784,720 (1991) and German Patent 4241045C1 (1990).
Rogner A et al. (1992) LIGA technique – What are the new opportunities. Journal of Micro-
mechanics and Microengineering, 2:133–140.
Rothschild M (2005) Projection optical lithography. Materials Today, 8:18–24.
Salib M, Michaeli A, Lazar A, Bechor H, Settle M, Krauss TF (2005) Photonic crystal fabrication
on silicon-on-insulator (SOI) using 248 nm and 193 nm lithography, Proceedings of SPIE –
The International Society for Optical Engineering, Nanophotonics for Communication:
Materials and Devices II, 6017:60170B.
Sayago I et al. (2005) Hydrogen sensors based on carbon nanotubes thin films. Synthetic Metals,
148:15–19.
Schaffer CB, Brodeur A, Garcia JF, Mazur E (2001) Micromachining bulk glass by use of femto-
second laser pulses with nanojoule energy. Optics Letters, 26(2):93–95.
Scheer HC, Schulz H, Lyebyedyev D (2001) Strategies for wafer-scale hot embossing lithography.
Proceedings of SPIE, 4349:86–89.
Schmidt RH, Haupt K (2005) Molecularly imprinted polymer films with binding properties
enhanced by the reaction-induced phase separation of a sacrificial polymeric porogen.
Chemistry of Materials, 17:1007–1016.
Sewell H et al. (2005) 32 nm node technology development using interference immersion lithog-
raphy. Progress in Biomedical Optics and Imaging – Proceedings of SPIE, Advances in Resist
Technology and Processing XXII, 5753:491–501.
Silverman PJ (2005) Extreme ultraviolet lithography: Overview and development status. Journal
of Microlithography, Microfabrication and Microsystems, 4:1–5.
Smardzewski RR et al. (2004) Metal-insulator-metal ensemble (MIME) chemical detectors. 2004
NSTI Nanotechnology Conference and Trade Show – NSTI Nanotech 2004, 3:163–164.
Smith BW et al. (2005) 25 nm immersion lithography at a 193 nm wavelength. Proceedings of
SPIE – The International Society for Optical Engineering, Optical Microlithography XVIII,
5754:141–147.
Solak HH (2006) Nanolithography with coherent extreme ultraviolet light. Journal of Physics
D: Applied Physics, 39:R171–R188.
Soltys J et al. (2003) Study of tip-induced Ti-film oxidation in atomic force microscopy contact
and non-contact mode. Acta Physica Polonica A, 103:553–558.
Steen SE et al. (2005) Looking into the crystal ball: Future device learning using hybrid E-beam
and optical lithography. Progress in Biomedical Optics and Imaging – Proceedings of SPIE,
Emerging Lithographic Technologies IX, 5751:26–34.
Steen S et al. (2006) Hybrid lithography: The marriage between optical and e-beam lithography.
A method to study process integration and device performance for advanced device nodes.
Microelectronic Engineering, 83:754–761.
Sun L et al. (2006) Preparation and gas-sensing property of a nanosized titania thin film towards
alcohol gases. Sensors and Actuators, B: Chemical, 114:387–391.
84 W.-C. Tian
Tachi S, Tsujimoto K, Arai S, Kure T (1991) Low temperature dry etching. Journal of Vacuum
Science and Technology B, 9:796–803.
Takeno T et al. (2005) Metal-containing diamond-like nanocomposite thin film for advanced
temperature sensors, Materials Science Forum. PRICM 5: The Fifth Pacific Rim International
Conference on Advanced Materials and Processing, 475–479:2079–2082.
Tanaka Y et al. (2005) Evaluation of pattern fidelity in EUVL using a high-numerical-aperture
small-field EUV exposure tool (HiNA). Progress in Biomedical Optics and Imaging –
Proceedings of SPIE, Emerging Lithographic Technologies IX, 5751:733–740.
Tian W-C, Pang SW (2001) Released submicrometer Si microstructures formed by one-step
dryetching. Journal of Vacuum Science and Technology B, 19:433–438.
Tian W-C, Pang SW (2002) Freestanding microheaters in Si with high aspect ratio microstruc-
tures. Journal of Vacuum Science and Technology B, 20:1008–1012.
Tian W-C, Pang SW (2003) Thick and thermally isolated Si microheaters for microfabricated
preconcentrators. Journal of Vacuum Science and Technology B, 21:274–279.
Tian W-C , Weigold JW , Pang SW (2000) Comparison of Cl 2 and F-based dry etching for highaspect
ratio Si Microstructures etched with an inductively coupled plasma source . Journal of
Vacuum Science and Technology B , 18 : 1890 – 1896 .
Ueno H et al. (1999) Study on fabrication of sub-micron structures for MEMS using deep X-ray
lithography. Proceedings of the International Symposium on Micro Machine and Human
Science, 87–92.
Vainos NA (2004) Laser grown photonic structures. Proceedings of SPIE – The International
Society for Optical Engineering, ROMOPTO 2003: Seventh Conference on Optics,
5581:1–11.
Weber W et al. (2005) Electron beam lithography for nanometer-scale planar double-gate transis-
tors. Microelectronic Engineering, Proceedings of the 30th International Conference on
Micro- and Nano-engineering, 78–79:206–211.
Wensink H, Berenschot JW, Jansen HV, Elwenspoek MC (2000) High resolution powder blast
micromachining. Proceedings of IEEE MEMS 2000, Miyazaki, Japan, pp. 769–774.
Yang X (2005) Hybrid exposure strategy: Combining e-beam direct writing with optical lithogra-
phy for magnetic recording heads. Journal of Vacuum Science and Technology B:
Microelectronics and Nanometer Structures, 23:2624–2630.
Yasin S et al. (2005) UVIII for combined e-beam and optical exposure hybrid lithography,
Microelectronic Engineering, Proceedings of the 30th International Conference on Micro- and
Nano-engineering, 78–79:47–50.
Zhao G et al. (2005) Direct electrochemistry and electrocatalysis of heme proteins immobilized
on self-assembled ZrO2 film. Electrochemistry Communications, 7:724–729.
Zhao Z et al. (2005) Annealing enhanced hydrogen absorption in nanocrystalline PdAu sensing
films. Journal of Applied Physics, 97:124301.
Zlobin VA (2005) Development of electron beam lithography for nanoscale devices, Proceedings
of SPIE – The International Society for Optical Engineering, Opto-Ireland 2005:
Nanotechnology and Nanophotonics, 5824:23–32.
Zribi A et al. (2005) Micromachined resonant multiple gas sensor. Sensors and Actuators,
A: Physical, 122:31–38.
Chapter 5
Applications of Functional Thin Films
and Nanostructures in Gas Sensing
Audrey Nelson
Abstract Thin-film technology plays a very important role in a variety of instru-

mentation used for gas sensing, such as optical nondispersive infrared (NDIR)
sensors. Because of recent developments in thin-film and nanostructure technolo-
gies, optical NDIR systems have become reliable and accurate instruments that are
both easy to use and economically viable. The author discusses various gas-sensing
technologies in general terms before narrowing the scope to infrared absorption
technologies and the significant role that functional thin films played in bring-
ing numerous infrared products to the market. The bulk of this chapter is used to
discuss the design and fabrication of an NDIR sensor as an example of a successful
product using functional optical thin films. The market for such sensors has grown
rapidly and more consumer products use them today.
Introduction
The BCC report (BCC research, 2005) C-245 from last year (Gas Sensors and Gas
Metering: Applications and Markets, by Edward Gobina, PhD, Published April
2005), reported that the global market for gas sensors and gas metering equipment
including secondary instrumentation was estimated at $2.8 billion in 2004 and
expected to rise at an average annual growth rate (AAGR) of 5.9% to $3.8 billion
in 2009. This rapidly growing industry has been pushing gas-sensing technology to
produce more reliable and more accurate instruments that are both easy to use and
economical. As these goals are met, the market may well open up and grow at even
faster rates.
Thin-film technology plays a very important role in a variety of instrumentation
used for gas sensing, including microelectromechanical (MEMS) gas sensors and
A. Nelson
GE Sensing, TelAire, 6860 Cortona Dr Suite B, Goleta, CA 93117-5568
e-mail: Audrey.Nelson@ge.com
86 A. Nelson
optical NDIR sensors, as well as in other sensors. MEMS technology is discussed

in another section. In this section we focus only on optical NDIR systems.
To understand the application of thin film in NDIR systems, we need to understand
how NDIR systems work. Therefore we start with the theory of light absorption,
and then discuss the nondispersive Mid-IR gas sensing systems. These discussions
portray the uses of thin-film technology in such systems. In the end, we use the
GE-Telaire NDIR system as an example, and discuss some infrared sources devel-
oped using thin-film technology.
Nondispersive Infrared (NDIR) Systems
Introduction
In a nondispersive infrared gas sensor system, a column of target gas absorbs

energy from a radiation source. Two fundamental processes are exploited to measure
the characteristics of the gas. The first process is caused by the energy absorption
in the gas, which results in a temperature rise and consequently an increase in gas
pressure. The second process is the attenuation of the incident radiation energy.
Two branches of gas-sensing instrumentation have developed based on these two
processes.
A system used to measure the pressure change in response to absorbed radiation
energy is called a photoacoustic system. The photoacoustic effect was discovered
around 1880 by Alexander Graham Bell, but only a century later was a modern
photoacoustic system developed by IBM scientists Tam and Patel, who used a laser
light source. A photoacoustic system has a light source, a resonant cavity filled with
the target gas called a photoacoustic cell, and a very sensitive microphone, which
sometimes is a piezoelectric crystal. The absorption of radiation energy by the gas
molecules causes an increase in the temperature of the gas, which in turn causes a
pressure increase that makes a pulse of sound which can be detected by the
extremely sensitive microphone that is located inside the photoacoustic cell. When
the intensity of the radiation source is modulated at the resonant frequency of the
cavity filled with target gas, the signal received by the microphone is amplified by
the quality or Q factor of the resonant cavity. The typical values of Q are between
10 and 1000, and thus photoacoustic systems can be extremely sensitive.
A variety of systems used to measure the attenuation of incident radiation energy
has been developed. Early systems dispersed the radiation into its separate wave-
length components with a grating. These dispersive spectrometers have a funda-
mental sensitivity limitation because one wavelength at a time is measured, and
almost all of the incident radiation is lost by the grating. The low level of light that
reaches the radiation detector suppresses the sensitivity of a dispersive spectrometer.
The basis of the modern nondispersive spectrometer, the interferometer, was pio-
neered by A. A. Michelson (Michelson Interferometer, 1882) and followed by
5 Applications of Functional Thin Films and Nanostructures 87
Charles Fabry and Alfred Perot (Fabry-Perot Interferometer, 1889). It took 60 years
for the advantages of using the Michelson interferometer as part of a nondis-
persive spectrometer to be recognized and nearly as many years for the Fabry-Perot
interferometer.
In the early 1950s, Peter Fellgett and Pierre Jacquinot employed the Michelson
interferometer to measure all wavelengths simultaneously, achieving an improve-
ment factor known as the multiplex or Fellgett’s advantage. In addition the interfero-
meter did not need slits to separate frequencies as the dispersive spectrometer does,
and there was also less reflection loss because there were fewer mirror surfaces;
considerably more energy reached the radiation detector. The higher level of light
achieved a second improvement factor known as the throughput or Jacquinot’s
advantage. These improvements are the basis of the Fourier transform spectrometer
(FTS), which enabled, for example, the first measurement of near-infrared emission
from the night sky by J. Connes in 1960 (Connes 1961). Dispersive and nondisper-
sive spectrometers are widely used in laboratories for the qualitative and quantita-
tive analysis of both gases and liquids. The use of spectrometers in industry has
been limited by the delicacy of the instrument.
A much simpler type of nondispersive spectrometer uses a noninterferometric
optical filter. These are generally called NDIR (nondispersive infrared) systems and
are used in the measurement of concentrations of gases that are of great interest to
industry such as CO2, CO, H2O, and CxHy. The multiplex advantage in an NDIR
system is actually greater than that in an FTS, because in an NDIR device all wave-
lengths arrive in phase at the radiation detector, and so there is no signal power loss
due to destructive interference. NDIR systems have a large throughput advantage
because there are fewer angular restrictions on the radiation when it is directed
through the sample gas column than there are in interferometers and dispersive
spectrometers. Last, but not least, a nondispersive system with an optical filter does
not need prisms, gratings, or interferometers, and therefore the optical engine can
be rather simple. Such a system can be both easy to use and economical, and thus
the demand for these devices has rapidly expanded in the past few years. We focus
our attention on nondispersive gas sensor systems using optical filters.
Theory of Absorption of Radiation
Radiation
Light or more formally electromagnetic radiation propagates through a vacuum at

a speed c = 3 × 108 m/s for all frequencies f. Consequently, the frequency and wave-
length l are related simply through the relationship f = c/l, and Fig. 5.1 shows the
wavelengths and frequencies of the electromagnetic radiation spectrum. Light
comes in packets of energy called photons. A photon is a particle of electromag-
netic radiation with energy E = h × f, where h is a universal constant of nature
88 A. Nelson
Fig. 5.1 Electromagnetic radiation spectrum
called “Planck’s constant” = 6.63 × 10−34 J s. Thus, the quantum energy of a photon
is also simply related to the inverse of the wavelength, with E (electron volts or eV)
= 1.24/l (microns or µm).
Each portion of the electromagnetic spectrum has quantum energies appropriate
for the excitation of certain types of physical processes. The energy levels for all
physical processes at the atomic and molecular levels are quantized, and if there are
available quantized energy levels in a particular form of matter that match the quan-
tum energy of the incident radiation, then the material will absorb that radiation,
and go to a higher energy state.
Gas molecules have a number of natural frequencies, which correspond to exci-
tations of quantized vibrations or rotations of the molecules. For a given molecule
and bonding structure, the natural frequencies and hence the photon energies that
may be absorbed are always the same. The magnitudes of absorption energies are
very low, in the tenths of electron volts, and so radiation with wavelengths in the
range of 2–20 µm (mid-IR range) are absorbed by molecules.
Absorption Spectra of Common Gases
Common gases such as water vapor (H2O), carbon dioxide (CO2), carbon monoxide
(CO), and hydrocarbons (CxHy), have many absorption lines in the mid-infrared
region. Fig. 5.2 shows the portion of the absorption spectra for some of these gases
between wavelengths of 4.2 and 5.6 µm. We see for each gas there are groups of
absorption lines known as vibration–rotation bands. For example, the absorption
spectrum of CO2 has a vibration–rotation band around 4.26 µm. The spacing
between distinct rotational states is generally smaller than the spacing between
vibrational states by approximately a factor of (m/M) 1/2 » 0.003, where m is the
mass of the electron, and M is the mass of the molecule.
Absorption Law
When infrared light passes through a gas, the gas has a high probability of absorb-
ing the light if the wavelength of the light matches a wavelength in the vibration–
rotation band.
Typical incandenscent lamp radiation intensity spectrum

1.2E-15
1E-15
Arbitrary units
8E-10
6E-16
Mid IR
4E-16
2E-16
0 Wavelength
0 1 2 3 4 5 6 7 8 9 10 in microns
Visible
100
Transmittance%
90
80
70 H2O
CO
60 N2O
50 CO2
40 Wavelength
4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 in microns
Transmittance spectra for N2O, CO, CO2,
and H20 vapor at 100 ppm-meters concentration
Fig. 5.2 Common gas absorption spectrum
The Beer–Lambert absorption law describes the reduction of the intensity I of a

monochromatic light beam as it passes through a small thickness dx of an absorbing
gas with concentration C. The change in intensity dI is related to other quantities
via the equation
dI / I = −aCdx, (5.1)
where a is the absorption coefficient of the specific gas in question at the specific
wavelength l of the incident light.
For a gas sample of the thickness L, the total change of intensity follows from
integration of the initial intensity I0 to final intensity I, which is equal to an integral
through the thickness L:
∫ dI / I = ∫ −aC dx → I + I 0 exp( −aCL )

(5.2)
Therefore transmission T
T = I / I 0 = exp( −aCL ). (5.3)
The transmission and the sensitivity to concentration are shown in Fig. 5.3 for a par-
ticular absorption column. When L = 0 or L = ∞, transmission has no sensitivity to
90 A. Nelson
1.2 0
T(L)
dT/dC(L) −0.000005
Sensitivity dT/dC measured in

1
−0.00001
0.8
−0.000015
Transmission
1/ppm
0.6 −0.00002
−0.000025
0.4
−0.00003
0.2
−0.000035
0 −0.00004
0 25 50 75 100 125 150 175 200
Path length in mm
Fig. 5.3 Transmission T and sensitivity dt/dc as function of path length
gas concentration, and between these extremes the magnitude of dT/dC peaks giv-
ing the optimal sensitivity when:
aCL = 1; L = 1 / (aC ). (5.4)
In practical applications, the light beam is not monochromatic, and the absorption
coefficient α varies as a function of wavelength. The light intensity transmitted
through a target gas of concentration C is then described by an integral of many
exponentials:
∫ i (l )e
( − a ( l )CL )
I (C ) = 0 dl , (5.5)
where i0 (l) is the intensity density per unit wavelength at wavelength l, and
i0 (l) = dI0/dl.
In principle, if we know α as a function of l, we can perform this integral.
In practice we can experimentally measure the ratio I(C)/I(0), where I(0) is the
transmitted intensity for zero target gas concentration, and I(C) is the transmitted
intensity as a function of target gas concentration C. I(C)/I(0) does not follow a
simple exponential as a function of C, but one can find a best numerical fit to the
experimental data to enable the extraction of C from a measurement of I(C)/I(0). In
some cases, I(0) may be estimated from an intensity measurement in a wavelength
band where α(l) is negligible. This technique allows I(0) to be measured simulta-
neously with I(C) in one gas volume.
Nondispersive Mid-IR Gas-Sensing Systems
Introduction
A nondispersive mid-IR gas-sensing system mainly has two parts: an optical

system and an electronic system. The optical system contains mainly a mid-IR
radiation source, a waveguide (often used not only to collimate the light beam, but
also as a gas sampling chamber), and an IR detector (often with filters if the IR
source is broadband). Most of these elements were developed using thin-film tech-
nology and we discuss them in more detail below.
A typical dual gas system is shown in Fig. 5.4.
Source and Waveguide
There are two types of mid-IR sources: quantum sources and thermal sources.
Quantum sources include lasers and LEDs. Early mid-IR laser sources are lead–
salt diode lasers (LSDL), made from IV–VI semiconductor materials, which operate
in the 3–30 mm spectral region. Because the LSDL requires a cryogenic cooling
system, it limits the range of uses. In recent years as the thin-film deposition tech-
nology underwent rapid development, many compact laser devices containing
antimonide compounds such as AlGaAsSb, InGaAsSb, and InAsSbP have been
developed for operation at higher temperatures. In 1994 Bell Labs developed the
first quantum cascade laser (QCL), which made a mid-IR room temperature operat-
ing laser possible (Faist et al. 1994; Gmachl 1998).
Because most laser sources are monochromatic, systems employing them do not
need an optical filter. The advantage of the laser source is high power (mW~W),
high sensitivity (ppb~ppt level), and very fast response time (~ns). However, most
Mirror Sample In Sample Out Mirror Chopper

Motor
Sample Cell
Reference Cell
Detector IR Source
Fig. 5.4 A NDIR system

92 A. Nelson
laser sources need cooling systems and are often complex to implement because of
stability issues. QCLs are not quite mature and are still too expensive for most
industrial applications. Mid-IR LEDs that do not require cooling are also scarce.
We discuss LED sources in a later section in more detail.
Thermal sources, such as incandescent lamps, are still most commonly used in
NDIR systems in industry. In general, thermal sources have two main shortcom-
ings: the first is their broadband radiation, which results in wasted energy because
only a narrow band is absorbed by the target gas; the second is their slow response
time which precludes fast modulation. For fast modulation with a thermal source,
one must use an optical chopper in the system configuration.
There are several reasons why modulation of the light source is desirable. The
most important one is that an alternating signal has a built-in zero reference. This
helps minimize the sensitivity of the offset drift and also is especially useful for
small AC signals in the presence of large DC offsets. Furthermore, filtering is easier
to implement into an AC signal by tuning the amplifier to pass the modulating
frequency and to reject all other frequencies, allowing one to reduce the noise level
considerably.
Because thin-film technology is undergoing rapid development, more and more
micromachined IR lamps and miniature blackbody sources (Axetris) have been
developed. They provide a mechanical stability advantage over lamps. Because the
source element can be made with membranes a few microns thick of very low ther-
mal mass, these sources often have fast response times.
The waveguide is part of the optical system, which may also include optical
mirrors or lenses to guide the light path. Simple waveguides can be made of either
metal or plastic with a few thin-film layers coating the inside of the waveguide.
Gold layer finishing is always desirable because gold has the best reflectivity for IR
light. The main purpose of the waveguide is to guide the radiation along the proper
path to the detector while minimizing the loss of radiation energy. In many simple
systems the waveguide is combined with a gas sampling chamber.
Detector
There are two types of mid-IR detectors: quantum detectors also known as
photodetectors, and thermal detectors. A quantum detector responds to individual
photons, which are the quanta of radiation. A thermal detector responds to tempera-
ture changes caused by the radiation power incident upon it.
There are many types of thermal detectors available, including thermopiles,
pyroelectric, and bolometer. The most desirable features of thermal detectors are
the high linearity of their signal as a function of the infrared radiation power and
their uniform response independent of wavelength.
A thermopile is an array of thermocouples connected in series. A thermocouple
is a pair of junctions of dissimilar metals that produce voltage when one side of the
junctions has a different temperature from the other. The so-called reference junc-
tion is kept at a known temperature by bonding it to a mass with a stable temperature.
The other, so-called active, junction is exposed to incident radiation. Connecting

many thermocouples in series increases the voltage and therefore allows greater
sensitivity. Some thermopile detectors have a built-in thermistor that measures the
reference temperature and thus allows the temperature of the target to be calculated.
The unique properties of the thermopile are its inherently stable response to DC
radiation and its relative insensitivity to ambient temperature variations.
A pyroelectric detector contains an active material such as lithium tantalite
(LiTaO3). When exposed to incident radiation, the resulting temperature change
causes a large electrical polarization of the active material. The electrical polariza-
tion induces a charge proportional to the incident power on nearby electrodes.
The advantage of the pyroelectric detector is its high responsivity: a pyroelectric
detector can output signals that are orders of magnitude higher than thermopiles,
which makes the pyroelectric detector easier to interface to external circuitry.
Additionally, the pyroelectric detector can be made with a fast response time close
to 1 ms, an order of magnitude faster than the thermopile. The disadvantage of the
pyroelectric detector is a strong sensitivity to ambient temperature and mechanical
vibration. Normally it has high noise. Recent data show that the detectivity (the
equivalent of the signal-to-noise ratio) of the pyroelectric detector and thermopiles
is similar (see Table 5.1).
A bolometer is a simple resistance thermometer. The responsive element changes
temperature when it absorbs incident radiation. The electrical resistance of the
responsive element changes due to the temperature change.
The performance for thermal detectors is limited by the minimum detectable
power or the noise equivalent power (NEP), which is a function of a mean square
fluctuation in radiated power, <D F2> (Wolfe WL, Zissis GJ (1985)) where
< DF 2 > = 4 kTd 2 GDf (5.6)
<D F> 2 − Mean square fluctuation in radiated power

k − Boltzmann’s constant
Td − Temperature of the detector
G − The thermal conductance between the responsive element and its
surroundings
Df − The electrical frequency bandwidth.
In general, for thermal detectors, all the quantities should be minimized to obtain
the smallest NEP. To optimize the detector sensitivity, materials with the lowest
thermal conductivity G should be chosen, and the narrowest bandwidth Df about the
frequency of the modulated signal should be selected.
A variety of quantum detectors is available for detection in the midinfrared. The
most popular ones are InSb, PbSe, and HgCdTe (CMT). All of these detectors
are extremely sensitive to temperature. Compared to performance at room tempera-
ture their detectivity improves by many orders of magnitude when these detectors
are operated at 77 K (liquid nitrogen temperature). Thus quantum detectors are
frequently used with cooling systems. These cooling systems increase the complexity,
size, and cost of the overall apparatus. Room temperature midinfrared quantum
94 A. Nelson
Table 5.1 Some Commonly Used Mid-IR Sources
λ
Life-time Modula- Vendor
Sources Range Photo
(h) tion
(µm)
International
Light
IR lamp 40,000 <2 Hz 0.2–5
Technologies
[ILT]
Nano-
Thermal Intex
amorphous 25,000 <100Hz 1 – 20
source [Intex]
carbon Films
Surface ICx
textured Photonics
25,000 <10 Hz 2 – 20
thin metal [ICx
filament Photonics]
Laser
Lead-salt
3 – 30 Components
Diode 10,0001 10kHz2
[Laser Com-
Lasers
ponents]
Roithner
Quantum LaserTech-
Laser Antimonide
2,0003 10KHz niK GmbH
Lasers
2–4 [Roithner
Laser]
Alpes
Quantum
Lasers
Cascade 7,000 10KHz [Alpes
Lasers 4 – 12 Lasers]
1
The Laser Guidebook by Jeff Hecht, Chapter 19, p. 280.
2
Laser sources are capable of modulation frequencies as high as a GHz, but for
gas-sensing uses, practical issues reduce the effective modulation frequency to the
kHz range.
3
“Midinfrared interband cascade lasers at thermoelectric cooler temperatures”
Electronics Letters Volume 42, 1034 (2006).
detectors present a great challenge, because of the limitations of the narrow band-
gap materials employed. Typical bandgaps are around ¼ eV, which is not too far
from the typical room temperature thermal energy kT ~ 0.025 eV. In Table 5.2, brief
characteristics of commonly available IR detectors and manufacturers are listed.
More detailed data such as normalized detectivity, time constant, resistance or
impedance, responsivity, noise spectrum, and spectral response are available from
manufacturers’ datasheets. One can choose the proper detector for a specific system
to match the required operating temperature, wavelength region, and other impor-
tant characteristics.
Table 5.2 Various Infrared Detectors

Spectral Operating Detectivity Da
Types Response (mm) Temperature (K) (CmHz1/2/W) Company
Thermo- Thermopiles 2–15 300 1.8 × 108 Perkin-Elmer
detectors (Perkin)
Pyroelectric 2–20 300 2 × 108 IGM&I Co. Ltd
(IGM&I)
Bolometer 8–14 300 1 × 109 Raytheon
(Raytheon)
Quantum InSb 2–6 213 2 × 109 Hamamatsu
detectors (Hamamatsu)
PbSe 1.5–5.8 300 1 × 108 Hamamatsu
HgCdTe 2–16 77 2 × 1010 Hamamatsu
As mentioned in an earlier section, photon energies in the infrared region are

intrinsically low. Therefore, the optimization of NDIR systems is crucial. The use
of electronic amplification is favored due to the rapid advances in electronics,
which have allowed high-performance electronic amplifiers to become available at
very low prices. Therefore, detector front-end electronic designs have become
easier. Many thermal detectors already have front-end electronics available inside
the detector package.
Filter
An NDIR system with a broadband radiation source often uses an optical narrow
bandpass filter, which is a component with a wavelength-dependent transmission or
reflectivity. The optical filter allows only wavelengths in the desired gas absorption
band to reach the detector. This increases the signal-to-noise ratio and minimizes
the effects of interfering gases, source variability at uninteresting wavelengths, and
broadband noise.
Although antireflection coating techniques had been developed in the early
1930s, and techniques for design of thin-film narrowband optical filters were well
established by the 1970s, the theory and techniques of layer thickness control were
apparently rather poorly developed. Over the last few decades, thin-film growth
technology and equipment have advanced dramatically. Now reliable narrow band-
pass filters with uniform response are available from many optical companies.
Depending on the shape of the transmission curve, one distinguishes filters of
the following types (Table 5.3).
There is a variety of designs of narrowband thin-film filters. The most common
type is based on the Fabry–Perot interferometer, which consists of two highly
reflecting mirrors with a half-wave dielectric layer in the middle, forming a
96 A. Nelson
Table 5.3 Different Filter Types
Types Transmission curve Definition
Bandpass Transmitting only a certain

filters wavelength range
Notch Eliminating only a certain wave-

Filters length range
Low Pass Transmitting only wavelengths be-

Filters low a certain value
High Pass Transmitting only wavelengths

Filters above a certain value
standing-wave cavity. Incident light passes through two coated reflecting surfaces.
The distance between the reflective coatings determines which wavelengths will
destructively interfere and which wavelengths will be allowed to pass through the
coated surfaces of the optical bandpass filters. In situations where the reflected
beams are in phase, the light will pass through the two reflective surfaces.
However, if the wavelengths are out of phase, destructive interference will block
most of the reflections, allowing almost nothing to be transmitted through the
optical bandpass filters. In this way, interference optical bandpass filters are able to
attenuate the intensity of transmitted light at wavelengths that are higher or lower
than desired. This structure can be made of metal-dielectric layers (where the
dielectric layer forms a spacer bounded by two metallic reflecting layers), all-die-
Table 5.4 Common Infrared Coating Materials (Harrick)

Wavelength Refractive
Material Range (mm) index@ (mm)
1.76 100
Percent transmission
80
60
Sapphire 0.1-5.0 40
20
0
1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 2.0 3.0 4.05.0 10 20
Wavelength in microns
1.47 100
80
60
BaF2 0.2-11 40
20
0
1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 2.0 3.0 4.05.0 10 20
3.42 100
80
60
Silicon 1.1-10 40
20
0
1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 2.0 3.0 4.0 5.0 10 20
100
2.49
80
60
40
ZnSe 0.6 – 20
20
0
1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 2.0 3.0 4.0 5.0 10 20
1.54 100
80
60
40
KBr 0.5 - 25
20
0
1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 2.0 3.0 4.0 5.0 10 20
lectric layers (a solid thin-film spacer with all dielectric multilayers on either side),
or a solid etalon plate (a cleaved crystal spacer with thin-film reflector deposited on
either side).
Most available midinfrared narrow bandpass filters are multiple cavity filters
that consist of a stack of static Fabry–Perot filters. A double cavity (structure such
as |reflector|half-wave|reflector|half-wave|reflector|) or triple cavity are very commonly
used in midinfrared narrow bandpass filters. The advantage of the multiple cavity
filters is a significant improvement in the shape of the transmission curve, the
desired wavelengths are transmitted with higher efficiency, and there is a sharper
and stronger cutoff of undesired wavelengths.
Many materials have been used to create thin-film coating and filter material.
Table 5.4 lists some of the common IR materials and their characteristics.
98 A. Nelson
GE-Telaire NDIR System
Introduction
Telaire (Telaire) is one of the world leaders in NDIR gas-sensing technology and is
a pioneer in the development of low-cost NDIR systems. Here I give a real example
of Telaire’s common NDIR system and then discuss mid-IR LED development
using high precision thin-film deposition molecular beam epitaxy (MBE).
Telaire Low-Cost NDIR System
Traditional NDIR systems typically involve a number of optical components

including an optical chopper, parabolic mirrors, multisample chambers, and so on.
Telaire’s design made the low-cost NDIR system possible. The system (see Fig.
5.5) includes four parts: IR source, waveguide (gas sampling chamber), detector,
and integrated electronics.
The source is an incandescent IR lamp with broadband emission; a narrowband
filter is mounted on the detector. The gas-sampling chamber is a waveguide,
designed using optical raytracing tools to minimize light loss.
This low-cost NDIR system has been vastly used in HVAC and in various
consumer products. The key operational characteristics of this system include:
● Dual beam detector: By using two detectors, each with a different narrow band-
pass filter, one tuned to the target gas absorbance and the other monitoring the
background, the system is able to correct for any light source and optical system
degradation bias.
Custom Designed
Infrared Filters
Reference Target Gas Dual Beam
Diffusion Membranes Micro-Machined
Thermopile Detector
Incandescent 1.25"
Infrared Source Patented Waveguide
Microprocessor
Fig. 5.5 Telaire NDIR system

• Lamp modulation: By modulating the lamp power the circuitry can use AC
signal processing to reduce any background radiation bias, detector drift, or
electronic drift.
• Waveguide: By making the gas-sampling chamber into an optical waveguide, it
is possible to both minimize light losses and optimize beam path length to fit the
gas concentration range. The longer the path length, the more molecules of
target gas will be in the path between the light detector and the source. Therefore,
low-concentration range sensors are designed with longer path lengths than
high-concentration range sensors.
• ABC logic: Crowning the above design is a patented (U.S. patents) Ambient
Background Calibration (ABC) tool for CO2 sensors. This calibration tool uses the
fact that the CO2 content in the atmosphere is quite stable to automatically correct
for any long-term drift by comparing the sensor readings with ambient back-
ground during any regular (e.g., daily) quiet periods where the sensor is used.
NDIR systems such as the one described have limitations. Because the lamp
consumes a significant amount of power it can’t be used in certain applications
requiring extremely low power. The lamp also has a slow response time, and so it
can’t be used when a fast response time is necessary. Furthermore, lamps have
stability issues. One solution is to replace the lamp with an infrared LED source.
IR LED development has a long history. A few pioneering groups in this field are
at the Russian Ioffe Institute and at Lancaster University in the United Kingdom.
There are frequently efforts reported in the international conference on Mid-Infrared
Optoelectronics: Materials and Devices (MIOMD).
Mid-IR LEDs and MBE
The fundamental difficulty of mid-IR semiconductor devices arises from the small
energy (in the tenths of electron volts) of mid-IR photons, so the appropriate materials
for this energy range have a small bandgap. Consequently Auger recombination,
where electron-hole pairs recombine and transfer their energy and momentum to
another carrier instead of emitting a photon, becomes important. At room temperature
and above, Auger recombination becomes dominant in the mid-IR. Small bandgap
materials also often have large spin-orbit interactions. In certain materials (e.g., InAs
and InSb) energy splitting due to spin-orbit interactions becomes comparable to the
bandgap itself, resulting in heavy intervalence band absorption.
These fundamental limitations have prolonged mid-IR room temperature LED
development for decades. So far, mid-IR LEDs are still commercially very scarce.
In order to develop a mid-IR LED that works at room temperature, Telaire has
explored some semiconductor heterostructure LEDs. The problem with these LEDs
is a spectrum shift due to temperature variation. Instrumentation is thus extremely
difficult because of the lack of stability over the desired temperature range. We
therefore put effort into the resonant cavity (RC) LED. The RCLED was first
100 A. Nelson
introduced and developed by Dr. Schubert in 1990 (Schubert et al. 1992). There are
many advantages of an RCLED compared with conventional LEDs, such as high
spectral purity, narrow angular distribution, high brightness, and high efficiency.
The emission line width of an RCLED is determined by the cavity quality fac-
tor Q, not by the thermal energy kT. Therefore, one can design ideal emission line
widths with respect to a target gas spectrum. At the same time, the RCLED has
much better temperature stability compared with conventional LEDs because the
wavelength shift with temperature is determined by the thermal expansion coef-
ficient of the optical cavity and not by the energy gap of the active material.
The performance of the device is critically dependent on two factors. The first is
the spectral overlap between the spontaneous emission peak from the active layers
and the cavity resonance, and the second is the match between the backside mirror
reflection and the cavity resonance. Therefore, high-precision growth techniques
are required. Molecular beam epitaxy (MBE) of thin-film deposition easily
provides the necessary precision.
MBE technology was developed by A. Y. Cho and his colleagues at Bell Labs
in late 1970s1994. “The unique feature of MBE is the ability to prepare single
crystal layers with atomic dimensional precision” (Alfred ).
An MBE is an epitaxial process where source material, such as gallium, indium,
and arsenic, are evaporated onto a heated crystal substrate inside an ultrahigh
vacuum chamber with pressure less than 10-6 torr. The vapors from the heated sources
form beams that travel through the chamber and land on the substrate surface in a slow
deposition rate (typical around 1 µm/h for most III-V materials), arranging themselves
in a crystalline lattice based on that of the substrate. During the operation, reflection
high-energy electron diffraction (RHEED) is used for monitoring the growth of the
crystal layers. A computer controls shutters in front of each furnace, allowing precise
control of the thickness of each layer, down to a single layer of atoms.
Telaire’s CO2 RCLED
So far most RCLED development has been in the visible and near IR region.
Mid-IR RCLED has certain challenges with its material system. However, for the
typical device with a few micron thicknesses, the cavity length fits the lowest orders
of the optical modes for mid-IR wavelengths. This promises a better overlap
between the resonant optical mode and the active region emission spectrum.
In 2004 Telaire developed a unique CO2 RCLED (Telaire RCLED) with an
emission spectrum that matches the CO2 absorption profile very well, as shown in
Fig. 5.6. Further narrowing of the LED spectral distribution can be achieved by
increasing the reflectivity of the elements of the resonant cavity.
Because of MBE’s ability for precise control, we have successfully grown our
RCLED structure1 to fit perfectly with the CO2 absorption spectral profile.
1
Hughes Research Laboratory performed the MBE growth of the RCLED structure.
1.8E-05
Ambient
N2 Purge
1.4E-05
Intensity (a.u.)
1.0E-05
6.0E-06
2.0E-06
3 3.5 4 4.5 5 5.5
Wavelength (um)
Fig. 5.6 RCLED spectrum
Table 5.5 Different System Configurations

Avg. Power
Source Detector Modulation Peak Current Consumption
IR lamp Thermopiles 1–5 Hz 100 mA 20–80 mW
LED Pyroelectric 0.25–1 kHz 100 mA 50–120 µW
LED Microphone 1–100 kHz 100 mA 1–100 µW
Compared with a conventional LED, the spectral distribution of our RCLED is

much purer. One of the major advantages of this device is that there is no significant
spectral shift over the temperature range 0-50 ©C or under different driving
currents (temperature data). Test data show that the RCLED offers more stable
performance than an incandescent lamp.
Summary
With the LED light source a variety of systems can be assembled. Table 5.5 shows
a comparison of the response time and power consumption of systems containing
an LED light source, as well as a typical system with an IR lamp for comparison.
In conclusion, thin-film technology has played a crucial role in NDIR systems.
Because of its fast development, one can imagine that many new systems will
emerge with faster and more reliable sources and detectors. As increasingly
102 A. Nelson
complex integrated circuits become available, sensors on the chip with smaller size
and better performance will soon become a possibility. Gas-sensing technology will
flourish with the many new NDIR devices.
References
Alfred C (1994) Molecular beam epitaxy. AIP, New York.

Alpes Lasers: http://www.alpeslasers.ch
Axetris: http://www.leister.com/axetris//sensors/irsource/index.html
BCC Research (formerly Business Communications Company, Inc): http://www.bccresearch.
com/instrum/C245.html
Connes J (1961) Rev. Opt. 40 45 l16 171 231.
Faist J, Capasso F, Sivco DL, Sirtori C, Hutchinson AL, Cho AY (1994) Quantum cascade laser.
Science 264:553–556.
Gmachl C (1998) Sensitive absorption spectroscopy with a room-temperature distributed-feed-
back quantum cascade laser. Optics Letters 23:219–221.
Hamamatsu: http://www.Hamamatsu.com
Harrick Scientific Products: http://www.harricksci.com/infoserver/optical%20materials.cfm
ICx Photonics: http://photonics.icxt.com/icx-photonics.asp
IGM&I Co. Ltd., St. Petersburg, Russia
ILT (International Light Technologies): http://www.gilway.com
Intex: http://www.eoc-inc.com/infrared_ir_pulsable_sources.htm
Laser Components IG. Inc: http://www.lasercomponents.com
MIOMD: The International Conference on Mid-infrared Optoelectronics - Materials & Devices
http://www.lancs.ac.uk/depts/spc/mirnet/miomd.htm
Perkin-Elmer: http://www.perkinelmer.com
Raytheon: http://www.raytheon.com
Roithner laser ThechniK GmbH: http://www.roithner-laser.com
Schubert EF, Wang YH, Cho AY, Tu LW, Zydzik GJ (1992) Resonant cavity light emitting diode.
Applied Physics Letters 60: 921–923.
Telaire: Telaire was founded by Dr. Jacob Y. Wong in 1990. Telaire is one of the first companies
to developing low cost NDIR systems for industry. The Director of Engineering Andrian
Kouznetsov is one of the pioneers in this field. Telaire holds more than 40 patents in gas sensor
technology. Their website is www.telaire.com
Telaire RCLED: This CO2 RCLED development was developed by Audrey Nelson, Principal
Scientist at Telaire, a unit of GE sensing. A patent application was filed in 2005.
Temperature data and driven current data are available in the patent application files.
U.S. Patents: 5,347,747 (1991); 6,526,801 (2000)
Wolfe WL, Zissis GJ (1985) The infrared handbook, 11–28 to 11–29.
Chapter 6
Chemical Sensors: New Ideas
for the Mature Field
Radislav A. Potyrailo
Abstract Chemical sensors for diverse applications for gas- and liquid-phase sensing
have their own design requirements. Thus, sensors typically have long timelines
from the concept through the evolution and cost reduction to commercial products.
For some applications, it is attractive to take advantage of previously developed,
optimized, and mass-produced physical transducers, optoelectronic, radiofrequency
identification, and other types of components and to rationally combine them with
sensing materials to produce new types of chemical sensors, more rapidly than it is
typically achieved. Widely deployed and accepted commodity consumer products
present a striking set of attractive capabilities applicable for advanced sensors.
This chapter presents several recent examples from our laboratory to demonstrate
developments in chemical sensors based on electrical, mechanical, and radiant signal-
transduction methodologies.
Introduction
Chemical sensors have found their niche among modern analytical instruments
when real-time determination of the concentration of specific sample constituents
is required. Development of sensors with new capabilities is driven by the ever-
expanding monitoring needs for determination of a wide variety of species in gases
and liquids. More sensitive sensors are required for analysis of ultratrace levels of
environmental pollutants, and pathogenic species in water. Sensors with faster
response are highly desired for monitoring of transient in vivo events and bedside
patients. More selective sensors are wanted for analysis of ambient urban or
battlefield air. For these and many other reasons, a wide variety of chemical sensors
and sensor systems is being developed.
R.A. Potyrailo
GE Global Research Center, 1 Research Circle, K1 3B39A, Niskayuna, NY 12309
e-mail: potyrailo@research.ge.com
104 R.A. Potyrailo
Based on a variety of definitions of sensors (Potyrailo 2006; Taylor and

Schultz 1996; Webster 1999; Wolfbeis 1991), here we accept that a chemical
sensor is an analytical device that utilizes a chemically responsive sensing layer
to continuously and reversibly recognize a change in single or multiple chemical
parameters of a measured environment and to convert this information into an
analytically useful signal. As shown in Fig. 6.1, in such device, a sensing mate-
rial is applied onto a suitable physical transducer to convert a change in a prop-
erty of a sensing material into a suitable form of energy. The obtained
signal from a single transducer or an array of transducers is further processed
to provide useful information about the identity and concentration of species in
the sample.
The energy-transduction principles that have been employed for chemical
sensing involve radiant, electrical, mechanical, and thermal types of energy
(Janata 1989; Middelhoek and Noorlag 1981/1982). Hyphenated techniques in
chemical sensing are well understood (Hirschfeld 1985; Hirschfeld et al. 1984)
and combine several transduction techniques in one sensor (Furuki and Pu 1992;
Potyrailo 2003). As shown in Fig. 6.1, in addition to a sensing material layer
and a transducer, a modern sensor system often incorporates other important
components such as sample introduction and data-processing components.
Examples of commercially available chemical sensor systems are presented in
Table 6.1.
Compared to chemical sensing based on intrinsic analyte properties (e.g., spec-
troscopic, dielectric, thermal, etc.), indirect sensing using a responsive material
expands the range of detected species, can improve sensor performance (e.g.,
analyte detection limits), and is more straightforwardly adaptable for miniaturiza-
tion (e.g., through MEMS or self-assembly). However, a possible drawback of the
indirect sensing approach is a trade-off between the selectivity of response to an
analyte of interest in a multicomponent complex sample and sensor reversibility.
Data processing
Transducer
Radiant energy
Sample Sensing Electrical energy Species identity
introduction material Mechanical energy Species concentration
Thermal energy
A B C D
Sample components
Fig. 6.1 Main components of a modern chemical sensor system

6
Table 6.1 Representative Examples of Commercially Available Chemical Sensor Systems

Type of Energy Involved
in Transduction Principle Type of Sensing Material Transducer Design Analyte/Applications Company
Electrical Composite conducting Array of inter-digitated Homeland security and Smiths Detection-Pasadena, Inc. (former
polymer electrodes defense applications Cyrano Sciences, Inc.), Pasadena,
CA, USA
www.smithsdetection.com
Electrical Nanoparticle metal oxide Micro-hotplate CO, NOx, O2, volatile Microchemical Systems SA, Corcelles,
organic compounds Switzerland
www.microchemical.com
Mechanical Polymeric film Array of surface acoustic Nerve and blister agents MSA, Pittsburgh, PA, USA
wave devices www.msanorthamerica.com
Mechanical Polymeric film Micro-machined strain Water vapor Hygrometrix Inc., Alpine, CA, USA
gauge www.hygrometrix.net
Chemical Sensors: New Ideas for the Mature Field
Radiant Polymeric film formulated Container-integrated pH, dissolved O2, pCO2 Fluorometrix Corp., Stow, MA, USA
with organic disposable patch www.fluorometrix.com
fluorescent dye
Radiant Metallo-porphyrin dye Colorimet-ric sensor Volatile organic compounds ChemSensing, Inc., Northbrook, IL,
on solid supports film array USA, www.chemsensing.com
Thermal Formulated catalytic Catalytic combustion Combustible hydrocarbons Delphian Corp., Northvale, NJ, USA
composition sensor www.delphian.com
Thermal Formulated catalytic Catalytic combustion Combustible higher Nemoto & Co, Milan, Italy
composition sensor hydrocarbons www.nemototech.com
105
106 R.A. Potyrailo
Requirements for Ideal and Practical Chemical Sensors
A wide variety of innovative ideas for sensors with new capabilities is being
originated by coupling new technological capabilities in micro- and nanofabrica-
tion of transducers (Hagleitner et al. 2002, 2003; Hierlemann and Baltes 2003;
Piner et al. 1999; Suzuki 2000) with new inorganic and polymeric sensing materials
(Adhikari and Majumdar 2004; Akmal and Usmani 1998; Frantzen et al.2004;
Janata and Josowicz 2002; Koinuma and Takeuchi 2004; McQuade et al.2000;
Potyrailo 2006; Scheidtmann et al.2005; Wolfbeis 2006) and with innovations in
sample manipulation (Erickson and Li 2004; Hansen et al.2002; Peterson 2005;
Tani et al.2004; Thorsen et al.2002; Vilkner et al.2004). Combinatorial and high-
throughput materials screening (Archibald et al. 2002a,b; Jandeleit et al.1999;
Koinuma and Takeuchi 2004; Maier et al.2002; Potyrailo and Amis 2003;
Potyrailo et al. 2004a; Potyrailo and Takeuchi 2005) and data-handling method-
ologies (Grate 2000; Jurs et al. 2000; Potyrailo et al.2004d) facilitate additional
improvements in sensor designs. Obviously, the design of a sensor for a particular
application will be dictated by the nature and requirements of that application.
Nevertheless, it is useful to set down the features that one would wish of an ideal
sensor for chemical species.
Ideally, the sensor would provide both adequate sensitivity (in some cases at the
single-molecule level) and a broad dynamic range. High selectivity towards the
species of interest and immunity to sample-matrix interferences should also exist.
In addition, an ideal sensor would be suitable for multicomponent measurements,
have fast and reversible response, and excellent long-term stability. Furthermore,
the ideal sensor must be robust, reliable, simple, economical to fabricate, of small
size, and with self-calibration capabilities. Of course, this “ultimate” sensor is far
from what is currently available. The most important respects in which existing
chemical sensors fall short are insufficient long-term stability, matrix interferences
(limited selectivity), and inadequate detection limits. Also, many modern analytical
sensing problems cannot be solved because of an absence of adequate detection
schemes for certain analytes and the practical challenges of distributed mapping of
analyte species over large remote areas or in vivo.
It is useful to note, that in real-world applications, the qualities of an ideal
sensor are often weighted differently according to application (Mitchell 1995;
Pickup and Alcock 1991). For example, reliability, long-term stability, and reso-
lution top the priority list for industrial sensor users whereas the size, cost, and
maturity of the technology are the least important factors. In contrast, medical
users focus on cost for disposable sensors (Mitchell 1995). The importance of
continuous monitoring also differs from application to application. For instance,
glucose sensing would still be acceptable if performed intermittently at a
frequency of 0.5–6 h whereas blood-gas sensors for use in intensive care should
be capable of continuous monitoring (Pickup and Alcock 1991). Specific
requirements for in vivo sensors include blood compatibility and sterilizability
(Meyerhoff 1993).
6 Chemical Sensors: New Ideas for the Mature Field 107
Numerous requirements for a practical sensor system (Ballantine et al. 1997;

Carrano et al. 2004; Harsanyi 1995; McQuade et al. 2000; Potyrailo et al. 1998b)
can be analyzed from the viewpoint of requirements for several key functionali-
ties and subsystems such as sample introduction, sensing materials, transduction
method and implementation, data analysis, and system packaging. As shown in
Table 6.2, the proper design of each individual subsystem has a tremendous
impact on the overall system performance. Interestingly, the sampling subsystem
seems to be affecting the majority of the performance characteristics of a successful
chemical sensor system. It is not surprising, because similar needs have been
reported for automated online process analytical instrumentation. For example,
the sample handling issues contribute 85% to the problems with UV/VIS process
analyzers whereas light source failures contribute 5%, and the remaining 10% is
distributed among electronics, detectors, and optical components (Potyrailo 2001;
Sherman 1996).
In designing a chemical sensor for gas- and liquid-phase sensing, attention
should be paid to specific requirements for these broad application categories.
Chemical sensors for gas- and liquid-phase chemical sensing have their own chal-
lenges and sensor-design aspects. Some such key challenges and sensor-design
aspects are summarized in Table 6.3.
Table 6.2 Field-Deployable Sensor Requirements

Sample Sensing Transduction
Requirements Introduction Material Principle Data Analysis
Dynamic range X X X X
Ergonomic design X
False positive rate X X X X
Initial cost X X X X
Long-term stability X X X X
Maintenance simplicity X X X
Multicomponent detection X X X X
Multiple operation modes X X X
Operation cost X X
Operation simplicity X X
Power consumption X X
Probability of detection X X X X
Response speed X X X
Response reversibility X X
Robustness X X X X
Selectivity X X X X
Self-calibration X X
Sensitivity X X X
Shelf-life X X
Size X X
Total (out of 20) 18 14 15 13
108
Table 6.3 Examples of Challenges and Sensor-Design Aspects Associated with Gas- and Liquid-Phase Chemical Sensing
Gas-Phase Chemical Sensing Liquid-Phase Chemical Sensing
Sub-systems Challenges Sensor-Design Aspects Challenges Sensor-Design Aspects
Sample – Particulate contamination – Preconcentration for detection limit – Corrosion – Protective coatings
introduction – Lack of representative improvement – Scale deposition – Replaceable sampling
sample for trace analysis – Lack of representa- components
– Time-domain modulation for
selectivity improvement – tive sample for trace
Integration into periodic self-clean- analysis
ing sensor system
Sensing material – Film poisoning – Temperature-stabilized operation – Reagent leaching – Covalent reagent attachment
– Water condensation (T > ambient) – Film poisoning to matrix
– Temperature, gas-flow modulation to – Film delamination – Cartridges of disposable sensors
facilitate more reversible and selective
response
Transducer – Corrosion – Transducer design for higher-order – Corrosion – Transducer design for higher-
– Signal-to-noise decrease with sensor response through hyphenated – Signal-to-noise order sensor response through
reduced transducer size transduction techniques and time-, decrease with reduced hyphenated transduction tech-
– Sophisticated readout with temperature-, sample- transducer size niques and time-, temperature-,
reduced transducer size flow-modulation methods for selec- – Sophisticated readout sample-flow-modulation methods
tivity, signal-to-noise, with reduced for selectivity, signal-to-noise,
stability improvement transducer size stability
improvement
Data – Reliable quantification of – Analysis of dynamic signatures for – Reliable quantification – Analysis of dynamic signatures
analysis four or less analytes in their selectivity and stability of four or less analytes for selectivity and stability
mixtures with sensor array improvement in their mixtures with improvement
of partially selective sensing – Analysis of multivariate signatures sensor array of partially – Analysis of mltivariate signa-
films and a single transduc- for selectivity, signal-to-noise, selective sensing films tures for selectivity, signal-to-
tion principle and stability improvement and a single transduc- noise, and stability
tion principle improvement
R.A. Potyrailo
Concepts for Ubiquitous Chemical Sensors
Existing sensors typically have long timelines from the concept through evolution
and cost reduction to commercial products. For example, micromachined pressure
sensors first were contemplated in the mid-1950s and became common in automo-
tive applications only by the late 1980s. Similarly, micromachined accelerometers
were invented in the mid-1970s, however, it was not until the early 1990s when they
became common in automotive applications (Hagleitner et al. 2002). Likewise,
chemical sensors take many years from the concept to a commercially successful
product (Wohltjen 2006). Thus it has been recognized that it is attractive to take advan-
tage of previously developed, optimized, mass-produced, and thus cost-effective
physical transducers (Boussaad and Tao 2003; Ren et al. 2005), optoelectronic (Cho
and Bright 2001; Cho et al. 2002; Ivanisevic et al. 2001; Manzano et al. 2003;
Vo-Dinh et al. 1999), radiofrequency identification (Finkenzeller 2003; Potyrailo
2006; Potyrailo and Morris 2007), and other types of components and to combine
them rationally with sensing materials to produce new types of chemical sensors,
more rapidly than is typically achieved. As an attractive complementary approach,
widely deployed and accepted commodity consumer products present a striking set
of capabilities applicable for advanced sensors (see Fig. 6.2)
Radiofrequency identification (RFID) tags have been recognized as one of the
disruptive technologies (Christensen 1997; Christensen and Raynor 2003) and are
widely used ranging from detection of unauthorized opening of containers, to auto-
matic identification of animals, to tracking insects, and to tagging of garments,
mailing labels, and combinatorial chemistry reaction products (Finkenzeller 2003;
Nicolaou et al. 1995; Wang et al. 2006). Part of the attractiveness of RFID tags is a
function of their low cost. For example, the cost of passive RFID tags that operate
at 13.56 and 915 MHz has dropped from $0.5–0.7 in 2003, to $0.04–0.4 in 2005,
and to $0.02–0.05 in 2006 (Bachner 2005; Lawrence 2005)
However, it was recently demonstrated that a multianalyte chemical identification
and quantitation can be performed using conventional RFID tags as transducers of
electrical energy (see Fig. 6.2a). Unlike other approaches where a special tag is
designed at a much higher cost (Mascaro et al.2007; Nambi et al.2003; Want 2004),
this approach utilizes a conventional RFID tag that is coated with a chemically sensi-
tive film. In such an RFID chemical sensor, both the digital tag ID and the complex
impedance of the antenna are measured. The measured digital ID can provide infor-
mation about the sensor and the object onto which the sensor is attached.
Highly efficient transducers of mechanical energy are also available as mass-
produced components. For example, it was demonstrated that micromachined
wristwatch tuning forks can be used as efficient chemical sensors when combined
with nanofabricated polymer wires for detection of a variety of vapors at part per
billion concentrations (Boussaad and Tao 2003; Ren et al.2005) as shown in Fig.
6.2b. Unlike the commonly used cantilever techniques, the fork converts the force
directly to an electrical signal without involving intermediate optical displacement
signal readout. This eliminates extra sources of noise, simplifies the instrumenta-
tion, and, allows a compact design for device applications.
110 R.A. Potyrailo
(a) 1 (b) prongs (c)
2 3
polymer wire
(d) illumination sequence 1 (e) (f)
illumination sequence 2
Fig. 6.2 Application of ubiquitous commodity electronic components and consumer products for
creation of chemical sensors. (a) Examples of 13.56 MHz passive RFID tags (with measured tag
type and tag ID) adapted for chemical sensing: (1) roll of RFID tags applicable for the roll-to-roll
deposition of chemically sensitive films, tag type = ISO 15693, ID of one of tags = E007 0000
0260 4AE8; (2) RFID tag on a plastic nonflexible support; tag type = I * CODE1, tag ID = 0900
000 3C8A 7AFA; (3) RFID tag on a plastic nonflexible support; tag type = I * CODE1, tag ID =
0900 000 457D 5E12. Scale bars are 10 mm. (b) Micromachined wristwatch tuning fork with an
attached nanofabricated polymer wire for chemical sensing. (c) Micromachined accelerometer
originally developed for automotive applications and coated with a sensing film for chemical sensing.
(d) Generated colorimetric patterns using a conventional computer screen for illumination of an
array of colorimetric sensing films. (e) Compact disk optical pickup head as a separate detection
unit outside an optical disk drive for analytical measurement. (f) Conventional optical disk drive
in a laptop computer for the quantitative chemical detection using chemical sensor films deposited
on conventional CD and DVD optical disks and extraction of analog signal from optical disk drive.
((a)–(f) are adapted from Lange et al. 2006, Manzano et al. 2003, Potyrailo et al. 2006, Potyrailo
and Morris 2007, Potyrailo et al. 2006, and Ren et al. 2005, respectively.)
Other attractive compact transducers of mechanical energy, also available as

mass-produced components, are micromachined accelerometers for automotive
applications. These transducers are attractive for chemical-sensing applications
because of several key reasons. First, these transducers have an integrated piezore-
sistive readout, which is already optimized and is very straightforward, providing a
desired reproducibility in measurements, while not requiring bulky equipment.
Second, chemically responsive film deposition is simple due to the ease of access
to the transducer’s surface. Third, the required sensitivity of the chemical sensor
can be achieved by choosing the right spring constant of the transducer. Thus, effi-
cient chemical sensors have been developed when chemically responsive layers
were applied onto these transducers (Potyrailo 2006) as shown in Fig. 6.2c.
Perhaps, the most diverse applications were demonstrated with transducers of
radiant energy that are available as mass-produced components. Optoelectronic
consumer products that are widely employed in the office and home are attracting
significant attention for optical sensor applications due to their cost advantage over
analytical instruments produced only in small quantities, robustness in operation
due to the detailed manufacturability improvements, and ease of operation (see
Fig.s 6.2d–f).
Computer screens have been applied for the illumination of colorimetric assays
(Filippini et al.2006; Manzano et al. 2003). Flatbed scanners were first applied to
analytical chemistry in the 1980s (Hruschka et al.1983) and continue to receive
attention today for sensor applications (Nath and Chilkoti 2002; Rakow and Suslick
2000; Taton et al.2000; Zhang and Suslick 2005). Handheld digital color analyzers
have been employed to quantitation of colorimetric sensor films (Hirayama et al.
2000; Suzuki et al. 2002). Compact disk optical pickup heads as separate detection
units outside optical disk drives have been used for a variety of measurement appli-
cations including scanning optical microscopy (Benschop and Rosmalen 1991),
position sensing (Chu and Lin 2005; Quercioli et al. 1999), and biodetection (Lange
et al. 2006). Applications of modified computer optical disk drives have been reported
for transmission measurements (Gordon 1999). Error determination routines in an
optical disk drive have been used to detect the presence of biological molecules and
bacteria on the disk surface (Jones 2005; Jones and Thigpen 2005; La Clair and
Burkart 2003). Finally, an attractive application of conventional computer optical
disk drives for chemical sensing has been demonstrated with a straightforward
extraction of an analog signal from a drive for the quantitative detection of optical
changes of chemical sensor films deposited on conventional CD and DVD optical
disks (Potyrailo 2006; Potyrailo 2006).
These examples conclusively demonstrate the power of implementation of existing
microfabricated, optoelectronic, and other consumer components to accelerate the
development of sensors with previously unavailable capabilities, functionality, or
level of acceptance.
Case Studies
Examples of recent reports from our laboratory demonstrate developments in the

most often used signal-transduction methodologies that include electrical, mechan-
ical, and radiant.
Electrical Energy Transduction
In sensors based on electrical energy transduction, sensing materials undergo

electrically detectable changes upon interaction with analyte species. Some examples of
measured electrical parameters include voltage, conductivity, complex impedance, work
function, capacitance, electrochemical potential difference, and current (Göpel 1996).
112 R.A. Potyrailo
Examples of sensing materials involved in these measurements include composite

polymers, conducting polymers, metal oxide semiconductors, and nanomaterials (such
as nanotubes, nanowires, etc.; Potyrailo 2006). Typical devices for these applications
include electrochemical and electronic transducers (Hagleitner et al. 2002, 2003;
Suzuki 2000; Wang 2002; Zemel 1990). The relative simplicity of fabrication of elec-
trodes makes sensors based on electrical energy transduction among the most
employed tools in chemical sensing. Typical electrode designs incorporate the electri-
cal transduction based on the changes in resistance, capacitance, or both.
Simultaneous measurement of resistance and capacitance changes can improve
sensor performance if these changes are not fully correlated. One of the approaches
to have a sensor with simultaneous measurement of resistance and capacitance
changes is to build a resonant LC circuit where both circuit inductance L and
capacitance C could be altered by the presence of analyte in a controlled manner.
It has been also understood that such resonant structures can be interrogated
remotely without a galvanic coupling but rather using an inductive coupling.
Proximity sensors operating on the principles of inductive coupling have been
under development since the 1950s when “endo-radiosonde” sensors of 2–6 mm in
diameter were reported (Mackay and Jaconson 1957) that consisted of a passive LC
resonant circuit, the resonant frequency of which was dependent on the pressure
surrounding the sensor and was measured with an external pickup coil.
We recently demonstrated a new approach for multianalyte chemical identifica-
tion and quantitation using conventional RFID tags as LC resonant circuits. Unlike
other approaches where a special tag is designed at a much higher cost (Mascaro
et al. 2007; Nambi et al. 2003; Want 2004), our approach utilizes a conventional
13.56 MHz RFID tag that is coated with a chemically sensitive film. A standard,
commercially available passive RFID tag coated with a sensing film does not need
a battery and includes a memory microchip connected to an antenna coil (Fig.
6.3a). The microchip can be read in a noncontact manner with an RFID reader by
illuminating the antenna (tuned by a combination of the antenna inductance (LA),
antenna capacitance (CA), and antenna resistance (RA)) with a RF carrier signal sent
by the reader (Fig. 6.3b). When the RF field passes through an antenna coil, ac
voltage is generated across the coil, which is rectified in the microchip to dc voltage
for microchip operation. The microchip becomes functional when the dc voltage
reaches a predetermined level. By detecting the RF signal backscattered from the
microchip, the information stored in the microchip can be fully identified. Upon
coating of the RFID tag with a chemically sensitive film, both the digital tag ID and
the complex impedance of the antenna are measured
Other reported RFID tags for sensing applications require a battery and/or a
specific redesign of portions of the electronic circuitry of the RFID tag (Artmann
1999; Mascaro et al. 2007; Nambi et al. 2003; Want 2004). Unfortunately, the most
prominent limitations of reported RFID and other proximity chemical, biological,
and physical sensors are difficulties with accurate measurements in the presence of
interferences, difficulties in quantitation of multiple parameters with a single sensor,
and the need for costly development of dedicated transducers for sensing.
(a)
Sensing film
RFID sensor
RF
CF Antenna
Sensing film
Inert substrate
(b)
Antenna
RA
RFID Micro-
LA CA
Reader chip
RFID tag
(c)
Fp
Zp
Zre
F2
Zim
F1
Frequency
Fig. 6.3 Strategy for the application of conventional passive RFID tags for chemical sensing. (a)
Adaptation of a conventional RFID tag for chemical sensing by deposition of a sensing film onto
the antenna. (Inset, analyte-induced changes in the film material affect film resistance (RF) and
capacitance (CF) between the antenna turns). (b) Schematic of digital ID reading from the RFID
tag and the equivalent circuit of the antenna of the RFID tag. (c) Measured parameters from a
single RFID sensor for multicomponent chemical detection and quantitation. Arrows indicate
changes of respective measured parameters
In our developed RFID sensor, the antenna coil not only completes its normal
activities, but is also tasked to act as an analytical electrode. This tasking of the
RFID antenna as an analytical electrode is possible because the complex impedance
of the RF resonant circuit formed by the antenna is sensitive to the dielectric
constant of any materials on or within the environment around the antenna.
114 R.A. Potyrailo
Consequently, if the dielectric constant of the local environment around the antenna
is adjusted, it will affect the complex impedance of the antenna circuit through the
changes in film resistance (RF) and capacitance (CF) between the antenna turns (Fig.
6.3a, inset). To provide for selective changes in the local dielectric field, a sensing
film must be applied to the antenna. Resultantly, only analytes that are selected by
the film will be incorporated into the local electrical environment and produce a
useful signal from the RFID sensor.
It is critical to match the response mechanism of the sensing film with the trans-
duction principle of the sensor. For selective detection of polar vapors with the
RFID sensor, we selected a solid polymer electrolyte film such as Nafion. This
polymer is of a family of perfluorosulfonated polymers, and is a copolymer of
tetrafluoroethylene and sulfonyl fluoride vinyl ether (see Scheme 6.1). Nafion is a
glassy polymer at room temperature with the glass transition temperature of its acid
form at ~110°C and of its salts at ~220°C (Yeo and Eisenberg 1977). The ionic
conductivity of Nafion and the possibility to absorb polar vapors are provided by
hydrophilic ionic groups (–SO3–H+) whereas hydrophobic “backbone” groups
(–CF2–CF2–) are relatively chemically inert (Morris and Sun 1993). The precise
mechanism of ionic transport in Nafion is still under investigation and possibly
involves the inverted micelle and the percolation effects (Tailoka et al. 2003).
The selectivity of response of Nafion to different vapors when Nafion is applied
onto the RFID sensor is provided by the differences in the resistance RF and capaci-
tance CF of the solid polymer electrolyte film on top of the RFID sensor antenna
structure upon exposure to different vapors (see Fig. 6.3a).
Scheme 6.1
Additional modifications of selectivity and sensitivity of Nafion sensing films on

RFID sensors can be easily accomplished by using different ionic forms of Nafion
(Kuban et al. 2004; Wang et al. 1997) and formulating Nafion with different materi-
als. Examples of formulation components in Nafion films include conducting poly-
mers (Smit et al. 2003), hydrogels (Madaras and Buck 1996), ionic liquids (Bennett
et al. 2004), salts (DeLongchamp and Hammond 2003), catalysts (Zen and Kumar
2001), surfactants (Singh and Shahi 1998), zeolites (Tricoli and Nannetti 2003),
nanowires (Wu et al. 2006), nanotubes (Su et al. 2006; Wang et al. 2003), sol–gels
(Feng et al. 1997), and some others. Because numerous formulation parameters
should be evaluated for the best performance of the sensor film, combinatorial and
high-throughput screening methodologies can be very useful in these optimization
efforts (Potyrailo 2006). In addition to Nafion, other materials with ionic conductiv-
ity can also be employed for multianalyte detection with a single RFID sensor.
Examples of such materials include poly(dimethyldiallylammonium chloride),

poly(vinyl chloride) formulated with tetrabutylammonium hexafluorophosphate,
poly(ethylene oxide) formulated with silver trifluoromethane sulphonate or lithium
perchlorate, and some others (Hughes et al. 2001; Opekar and Štulik 1999).
For multicomponent detection and quantitation using a single RFID sensor,
multiple parameters from the measured real and imaginary portions of the com-
plex impedance were calculated (Fig. 6.3c). These parameters included maxi-
mum frequency of the real part of the complex impedance (Fp), magnitude of the
real part of the complex impedance (Zp), resonant frequency of the imaginary
part of the complex impedance (F1), and antiresonant frequency of the imagi-
nary part of the complex impedance (F2). Because F1 and F2 are related to dif-
ferent components of the equivalent circuit, they are not expected to be
correlated. Examples of passive RFID tags modified for chemical sensing are
presented in Fig. 6.2a.
Effects of different vapors on the response of the RFID sensor were evaluated
by selecting a difficult combination of vapors such as methanol (MeOH), ethanol
(EtOH), water (H2O), and acetonitrile (ACN). Water vapor served as a control vapor
and acetonitrile was selected as a simulant for blood chemical warfare agents
(CWAs). Differences in the combination of the analyte and properties of the Nafion
sensing film affected the response of the RFID sensor to these different vapors. Fig.
6.4a demonstrates the measured Zp response for four analytes (H2O, EtOH, MeOH,
and ACN) for multiple concentrations and replicates (n = 3). Measurements of a
single parameter of an RFID sensor, for example, Zp, cannot discriminate between
different analytes. For example, if a signal Zp is changed by ~20 ohm, this change
can be due to 0.1 P/Po of H2O or 0.15 P/Po of MeOH or 0.2 P/Po of EtOH. Thus, a
single-parameter measurement of the RFID sensor cannot discriminate between
different analytes and their concentrations. Thus, to provide selective response from
the RFID sensor, multivariate analysis, for example, principal components analysis
(PCA; Beebe et al. 1998), can be used to extract the desired selectivity response
pattern from the data
Results of the PCA multivariate analysis of multiparameter (F1, F2, Fp, and Zp)
response of the RFID sensor to the changes in H2O, EtOH, MeOH, and ACN for
six analyte concentrations each are presented in Fig. 6.4b. For PCA, values of F1,
F2, Fp, and Zp from each of three replicate exposures to respective analyte concen-
trations were used. Prior to PCA, data were appropriately preprocessed by autoscaling.
The scores plot of the first two principal components of data shows a complex rela-
tion between measured signals from one sensor. Fig. 6.4b illustrates that three out
of four vapors are resolved using a selected sensing film and the complex imped-
ance readout from the RFID sensor. In particular, H2O vapor is well discriminated
from ACN and MeOH or EtOH vapors. However, MeOH and EtOH vapors are not
appreciably discriminated.
We further evaluated initial signal stability and the detection limit of vapor deter-
minations with RFID sensors. In this experiment, measurements of RFID sensor
response to different concentrations of water vapor were performed over 9 h.
The detection limit was calculated at S/N = 3 from the slope of sensor response to
116 R.A. Potyrailo
10 (a)
−10
Delta Zp (Ohm)
−20
−30
H2O
−40
MeOH
−50 EtOH
ACN
−60
−70
0 0.5 1 1.5 2 2.5 3 3.5
Time (h)
1.5
(b)
MeOH
1 ACN 0.2
0.2EtOH 0.15
0.15 0.2
0.5 0.1
0.1
0.15 0.07 0.07 0.1
0.15
0 0.04 0.07
0.1 0.04
PC 2
0.04
0.02 0.02
−0.5 0.07 0.02
0.04
0.02
0
−1
H2O
−1.5 0.2
−2
−2 −1 0 1 2 3 4 5 6 7 8
PC 1
Fig. 6.4 RFID sensor response. (a) Measured Zp response for four analytes (H2O, EtOH, MeOH,
and ACN), six concentrations (0, 0.02, 0.04, 0.07, 0.10, 0.15, and 0.20 P/Po) and three replicates.
(b) Results of the PCA multivariate analysis of measured parameters F1, F2, Fp, and Zp of the
RFID sensor to the changes in H2O, EtOH, MeOH, and ACN at six concentrations each, and three
replicates per concentration. Numbers are P/Po values for individual vapors
the smallest measured vapor concentration (0.02 P/P0) to be 600 ppb. Upon a care-
ful selection of the sensing material for a specific application, such low detection
limits obtained with the RFID sensors were attractive for diverse applications in
environmental, homeland security, and other demanding applications.
Critical to the wide acceptance of the demonstrated RFID sensors is the
analyte-quantitation ability of these sensors without repositioning errors between
the RFID sensor and a pickup coil. We evaluated the capability for position-
independent (RFID sensor: pickup coil) analyte quantification using multivariate
analysis tools. For these evaluations, we assembled a test system that contained
an RFID sensor positioned in a low dead volume gas flow cell as shown in Fig.
6.5a. A pickup coil was positioned outside the flow cell on an X–Z translation
stage. Thus, the relative position of the RFID sensor and the pickup coil was
changed in a controlled fashion. The studied repositioning conditions are illus-
trated in Fig. 6.5b and included 5 mm step changes in the Z-direction
(0, 5, 10, 15, 20 mm, and back to 0 mm) and 5 mm step changes in the X-direction
(0, 5, 10, 15, 20, 15, 10, 5, 0 mm). To perform analyte-quantitation measure-
ments, the RFID sensor was coated with a composite poly(vinyl acetate)/carbon
black polymer film responsive to water vapor. For each position of the sensor,
two replicate exposures to water vapor were performed. Thus, during each step,
the sensor environment was switched twice from dry nitrogen to 45% relative
humidity (RH) nitrogen. The sensing film demonstrated a desired completely
reversible response (see Fig. 6.5c). Upon sensor repositioning, we were inter-
ested in three aspects of position-induced sensor response effects: (1) sensor
signal upon exposure to dry carrier gas (baseline signal), (2) sensor signal
change from the baseline signal upon exposure to 45% RH (analyte response),
and (3) sensor noise.
Results of the measured response of the RFID sensor to the changes in RH at
different distances from the pickup antenna are summarized in Fig. 6.6a–d. The
measured parameters included parameters from the real and imaginary portions
of the complex impedance, F1 shift, F2 shift, Fp shift, and Zp, as described
above. Unfortunately, as shown in Fig. 6.6a–d, none of the individual measured
parameters provided a position-independent analyte quantitation.
The position-induced affects can be summarized as the changes in the baseline
sensor signal and changes in analyte response magnitude upon exposure to 45%
RH. It was also observed that the sensor noise did not significantly change under
these testing conditions.
To address this problem in significant sensitivity of the sensor response to the posi-
tion of the pickup coil, we performed a principal components analysis of measured
parameters. The goal was to identify a possibility of rejecting the repositioning effects
in the multivariate domain. Results of the multivariate analysis of response of the RFID
sensor to the changes in RH at different distances from the pickup coil are summarized
in Fig. 6.6e. Thus, using multivariate analysis of RFID sensor response, we were able
to preserve position stability of the baseline sensor signal upon exposure to dry carrier
gas and the magnitude of the sensor signal change upon exposure to 45% RH.
118 R.A. Potyrailo
(a)
RFID sensor tag
Z
Gas flow cell X
Network
analyzer
RF pickup antenna
(b)
25
Z Direction X Direction
20
Distance (mm)
15
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Positions
(c)
RFID Sensor Signal
45%RH 45%RH
(Arb. Units)
5 min
Time
Fig. 6.5 Evaluation of the analyte-quantitation ability of RFID sensors in the presence of repositioning
errors between the RFID sensor and a pickup coil. (a) Assembled test system. (b) Different tested posi-
tions of the RFID sensor in the X- and Z-directions with respect to the pickup antenna. (c) Replicate
(n = 2) response of the RFID sensor to 45% RH and dry N2 for each X and Z position of the sensor
We also explored the applicability of the RFID sensors for chemical analysis in
liquids. Of course, the resonance of the sensor will be dumped in high-conductivity
solutions. However, measurements can be made in low-conductivity solutions, such
as deionized water. An isolating layer can be applied onto the sensor to make sensor
(a) (b)
150000 20000
Frequency Shift (Hz)

F1 Shift (Hz) F2 Shift (Hz)

0
100000 −20000
−40000
50000
−60000
−80000
0
−100000
−50000 −120000
0 80 160 240 320 400 480 0 80 160 240 320 400 480
Time (min) Time (min)
(c) (d)
40000
Fp Shift (Hz) Zp
150
20000
0
100
Zp
−20000
−40000 50
−60000
0 80 160 240 320 400 480 0 80 160 240 320 400 480
Time (min) Time (min)
(e)
5
3
Principal Component #2
−1
−2
−3
−4
−5
0 500 1000 1500 2000 2500 3000 3500 4000
Sample Number
Fig. 6.6 Multivariate analysis provides position-independent analyte quantification. (a) F1 fre-
quency shift of the maximum of the imaginary part of the complex impedance; (b) F2 frequency
shift of the minimum of the imaginary part of the complex impedance; (c) Fp frequency shift of
the maximum of the real part of the complex impedance; (d) Zp, the magnitude of the real part of
the complex impedance; (e) Results of principal components analysis of measured parameters
120 R.A. Potyrailo
(a)
100
0 Ion conc. (ppb)
99 4
8
98 12
Zp (Ohm)
100
97 20
98
Zp (Ohm)
96 96 28
95 94
92
94 0 10 20 30 40 50 44
NaCl conc. (ppb)
0 200 400 600 800 1000

Time (s)
(b)
1.2
bare
1 sensing film
Zp normalized
0.8
0.6
0.4
0.2
0
NaSO4 HCl NaCl KH2PO4
(c) Ionic Species
60000
50000
Freq. shift (Hz)
40000
30000
20000
10000
0
0 2 4 6 8 10
Acetone conc. %vol.
Fig. 6.7 Applicability of RFID sensors for monitoring in liquids. (a) Analysis of ion concentra-
tions in deionized water. Inset, linear calibration curve of detection of NaCl in deionized water
over ion concentration range from 0 to 40 ppb. (b) Response patterns of bare and polysulfone-
coated RFID sensors upon exposures to different ions in deionized water. (c) Quantitation of
residual organic solvents in water using a polysiloxane-coated RFID sensor
operation possible in solutions with high conductivity (Hofmann et al. 2005; Ong
et al. 2001). Analysis of concentrations of ions in water was performed with such a
sensor as shown in Fig. 6.7a. While the sensor was in contact with deionized water,
increasing concentrations of NaCl were added to the water. As expected, the sensor
response was affected by the increasing ion concentration. The inset of Fig. 6.6a
shows the linear calibration curve of detection of NaCl in deionized water over ion
concentration range from 0 to 40 ppb. The calculated detection limit (at S/N = 3) was
460 part per trillion (ppt). Because these measurements were performed with an
uncoated RFID sensor, we effectively measured water conductivity change related to
NaCl concentration. Of course, coating of the sensor with a chemically sensitive film
should provide a selective response to different ions in water. Fig. 6.7b illustrates the
change in the response pattern of the RFID sensor response upon coating of the sen-
sors with a polysulfone sensing film and analysis of different ionic species such as
NaSO4, HCl, NaCl, and KH2PO4 in water. Furthermore, another RFID sensor was
coated with a hydrophobic sensing film (polysiloxane) to demonstrate quantitation of
residual organic solvents in water. In particular, Fig. 6.7c illustrates a response of an
RFID sensor upon interactions with different concentrations of acetone in water.
In conclusion, we have shown that a careful design of the electrical energy trans-
ducer to measure several parameters at once provides a desired improvement in
sensor selectivity.
Mechanical Energy Transduction
In sensors based on mechanical energy transduction, sensing materials undergo changes

in their viscoelastic properties or mass. The most widely used sensing materials for this
type of sensors are polymeric and copolymeric materials. Often, these sensors are used
for detection of gases and vapors with the polymer–analyte interaction mechanisms that
include dispersion, dipole induction, dipole orientation, and hydrogen bonding interac-
tions (Grate 2000; Grate et al. 1997b). Other materials include inorganic vapor-sorbing
materials such as zeolites, nanotubes, nanowires, ionic liquids, fullerene, graphite, and
others (Potyrailo 2006). The mechanical energy type of transduction has been employed
for chemical sensing since the mid-1960s (King 1964). At present, typical devices for
these applications utilize acoustic-wave resonant (Ballantine et al. 1997; Potyrailo et al.
2003; Thompson and Stone 1997; Ward and Buttry 1990) and cantilever transducers
(Lavrik et al. 2004; Sepaniak et al. 2002).
Employing such acoustic-wave resonant transducers, we recently have demon-
strated new sensing materials based on silicone block polyimide polymers (see
Scheme 6.2) for detection of volatile organic compounds, such as chlorinated
O O O O
N N N N Si-O Si O -Si
O O O O
O O O O y
x ~10
Scheme 6.2
122 R.A. Potyrailo
40 (a)
5.3
Sensor response (− Hz)

30
2.4
20
1.5
10
0.5
0
0 200 400 600 800 1000
Vapor concentration (ppb)
5.5
(b) y = 4.818−0.493x R2 = 0.999
5
4.5
log (K)
3.5
3
conventional polymers
2.5
−2 −1 0 1 2 3
Log (Vapor concentration, ppm)
(c) O C
Signal Intensity (counts)
Three-years old N Si
Fresh
1000 800 600 400 200 0

Binding Energy (eV)
Fig. 6.8 Performance of silicone polyimide as a vapor-sensing material upon deposition onto a
10 MHz TSM resonator: (a) concentration response curves for the determination of part-per-bil-
lion levels of TCE with 0.5–5.3 mm thick sensing films. (b). Partition coefficient as a function of
TCE concentration. (c) Long-term stability of new sensing material. Comparison of oxidation of
the freshly coated and three-year-old sensing films as determined by XPS
organic solvent vapors at trace levels (Potyrailo and Sivavec 2004). Typical calibration
curves for the determination of part-per-billion levels of TCE for several films of
different thickness deposited onto 10-MHz thickness-shear mode resonators are
presented in Fig. 6.8a. We observed that the sorption isotherm of new sensor poly-
mers to different chlorinated and other types of vapors was nonlinear.
The origin of this nonlinearity is likely to be from a combination of different vapor-
sorption mechanisms with contributions from both Langmuir- and non-Langmuir-
type noncovalent vapor–polymer interactions. When the frequency change was
related to the square root of the vapor concentration, linear responses were observed
with a slight upward non-linearity with the thickest tested film.
The magnitude of the response of conventional polymer materials at these concen-
trations was negligible, <1 Hz because partition coefficients of different conventional
polymer materials for TCE and other chlorinated volatile organic vapors of environ-
mental interest are typically (0.5–2) × 103 as reported in the literature (Finklea et al.
1998; Grate et al. 1997b, 1998; Jarrett and Finklea 1999; Patrash and Zellers 1993).
These partition coefficients are calculated assuming the sensor response is due to
mass-loading only (Grate et al. 1997b). Fig. 6.8b compares the partition coefficients
for TCE of new and conventional polymers. These data demonstrate a more than 100-
fold enhancement of the partition coefficient of our new material for detection at ppb
levels when compared to conventional currently used sensor materials such as phenyl-
methyl-polysiloxanes, poly(epichlorohydrin), poly(isobutylene), poly(ethylene
maleate), poly(ethylenimine), and many others (Bender et al. 2003; Grate et al.
1997a, 1998; Patrash and Zellers 1993, 1998; Rosler et al. 1998).
The concentration dependence of the partition coefficient as seen in Fig. 6.8b
originates from the nonlinear response of the materials to vapors. We believe that
this high partition coefficient may originate from the very favorable solubility of
organic vapors in the combination of the hard and soft blocks of this sensor mate-
rial. It is likely that in our silicone block polyimide, the generated microphase
domains with the approximate size of the polymer soft and hard segments create
sites for very favorable solubility interactions with the vapors.
Long-term stability of sensor materials is the critical parameter for practical
sensor operation. Oxidation of a sensor polymer is one of the main degradation
mechanisms that should be avoided. Oxidation may lead to the decrease of material
response to nonpolar organic vapors and to the increase of the response to polar
vapors, with water vapor being the main concern. We studied oxidation of freshly
coated and three-year-old sensor films using X-ray photoelectron spectroscopy
(XPS). No detectable differences in the amount and types of oxygen on the film
surface were found as shown in Fig. 6.8c. We also have found that the film thick-
ness was stable to within 5% as measured by the value of fundamental frequency
of coated crystals over time. This long-term stability of our new sensor materials is
provided by the nature of silicone polyimides. The flexible polysiloxane segments
enhance the stability of imide segments, and improve adhesion and membrane
properties of the film (Chang et al. 2001; McGrath et al. 1999). We also observed
that unlike conventional polyimides, water sensitivity of the new hybrid polyimides
was suppressed because of the silicone soft block.
The high sensitivity and long-term stability of these sensor materials make them
attractive for ultrasensitive sensors. These materials were implemented as thin films to
detect part-per-billion concentrations of TCE with 3 ppb limit of detection. Detection
limits for other chlorinated organic solvent vapors such as perchloroethylene (PCE),
124 R.A. Potyrailo
cis-1,2-dichloroethylene (DCE), trans-1,2-DCE, 1,1-DCE, and vinyl chloride (VC)

were 0.6, 6, 6, 11, and 13 ppb, respectively. We have also demonstrated a four-sensor
array with amorphous fluoropolymer (Potyrailo 2002), silicone polyimide (Sivavec and
Potyrailo 2002), poly(isobutylene), and poly(vinyl propionate) sensing films that selec-
tively and quantitatively determined TCE, three isomers of dichloroethylene (DCE),
and vinyl chloride (VC) (Potyrailo et al. 2004b; see Fig. 6.9). The high sensitivity and
good long-term stability of these discovered sensing materials make them promising
candidates for the field applications. An example of our recently performed initial field
tests (Shaffer et al. 2003) with several kinds of these silicone polyimides developed
using high-throughput screening (Potyrailo et al. 2004c) is shown in Fig. 6.10.
In our other recent application, vapor-sorption effects were also detected using
micromachined physical transducers such as cantilevers with integrated piezoresis-
tive readout. Often, cantilever response is measured in air on an atomic force micro-
scope (Su et al. 2003) or using an optical deflection readout (Dutta et al. 2003;
Hansen et al. 2001; Savran et al. 2004). These costly, complicated, and bulky rea-
dout schemes make the whole sensor systems difficult to miniaturize and preclude
them from being efficiently integrated into low-cost and small-size packaged sensor
suites. In our approach, although originally these transducers have been developed
and optimized as sensitive accelerometers for automotive applications, we applied
a chemically responsive layer onto the seismic mass of the transducer and con-
verted these transducers into chemical sensors (see Fig. 6.2c).
The performance of chemical sensors was evaluated in detection of carbon dioxide.
For carbon dioxide detection, a polycarbonate polymer sensing film has been
deposited into the cantilever. Upon deposition of a polycarbonate sensing film, the
4 1
2
Scores on PC# 2
0
3
-2
4 5
-4
2
-6
-8
-10
0 10 20 30 40 50 60 70 80 90 100
Scores on PC# 1
Fig. 6.9 Plot of the first two principal components for the dataset obtained from the response of
the four-sensor array to five chlorinated organic vapors at different concentrations. Vapors: 1, cis-1,
2-DCE; 2, TCE, 3, trans-1, 2-DCE, 4, VC; 5, 1, 1-DCE. Each data point is the mean of three
measurements; squares represent one standard deviation
Fig. 6.10 Field-testing of a sensor system with several kinds of silicone polyimides developed
using high-throughput screening techniques
sensor’s response to carbon dioxide was linearly proportional to the polymer film
thickness. Polycarbonate polymer is a glassy polymer and has been employed
earlier for the reversible CO2 detection at room temperature (Ahuja et al. 1999).
The mechanism of the polymer response to CO2 is based on the diffusion of the gas
into the “microvoids” of the polymer film resulting in an increase of the film mass.
This polymer–CO2 interaction causes no detectable change in the mechanical prop-
erties of the film and minimal relaxation of the polymer matrix (Ahuja et al. 1999).
Upon CO2 sorption into polycarbonate, the mass increase of the film was detected
with the transducer. Other materials for CO2 determinations at room temperature
include tetramethylammonium fluoride tetrahydrate (Gomes et al. 1996), aminopro-
pylsiloxane–octadecylsiloxane copolymer (Oprea et al. 1999), polyethyleneimine
(Korsah et al. 1998), fluorinated polyimide (Hoyt et al. 1998), and tetrakis(hydroxy-
ethyl)ethylenedlamine (Fatibello-Filho et al. 1989).
Reproducibility and hysteresis-free sensor response are among the key figures
of merit of a chemical sensor. These metrics depend on the performance of both the
transducer and sensing film. Reproducibility of the sensor response was evaluated
by repetitively exposing the sensor to different CO2 concentrations in both increas-
ing and decreasing concentration sequences. We intentionally selected high CO2
concentrations (10–100% vol. CO2) to provide the worst-case scenario for the
sensor operation to study hysteresis effects. Quantification of sensor hysteresis was
performed over the wide concentration range as shown in Fig. 6.11a. We have
found that the response of the sensor to increasing and decreasing concentrations
of CO2 had no detectable hysteresis effect. The sensitivity of the CO2 sensor operat-
ing at an excitation voltage of 7.5 V was 57.50 ± 0.38 mV/1% vol. CO2. At these
testing conditions, the sensor demonstrated an excellent reproducibility. Repetitive
exposures to 100% vol. CO2 provided signal reproducibility of 0.17% RSD. From
126 R.A. Potyrailo
6400
5600
(a)
Sensor Response (mV)

4800
4000
3200
2400
1600
CO2 increase
800 CO2 decrease
0
0 20 40 60 80 100
CO2 Concentration (%)
4000
(b)
3000
2000
1000
0
−1000
−2000
−3000
0 2000 4000 6000 8000
Time (s)
250
(c)
200
150
50
100
25
50
0
0 4000 8000
0
0 10000 20000 30000 40000
CO2 Concentration (ppm)
Fig. 6.11 Performance of polycarbonate sensing film upon deposition onto a micromachined
cantilever with an integrated piezoresistive readout. (a) Quantification of sensor hysteresis. (b)
Reproducibility of sensor response evaluated by repetitively exposing sensor to different (0%,
10%, 30%, 50%, 75%, and 100% vol.) CO2 concentrations. (c) Calibration results for the CO2
sensor for part-per-million CO2 concentrations
the analysis of the dynamic data collected during the reproducibility studies (see
Fig. 6.11b), we evaluated the response time of the sensor to a chemical stimulus.
We measured the time T90 required to achieve a 90% signal change upon a step
change of CO2 gas concentration to be about 45 s.
Evaluation of the sensor response to part-per-million concentrations of CO2 was

performed to determine the detection limit of the sensor. At low concentrations of CO2,
the sensor had a sensitivity of response of 5.63 mV/1000 ppm of CO2 as determined
from the fit of the response over the 0–8000 ppm concentration range; see Fig. 6.11c.
This sensitivity was very similar to that obtained for high concentrations. The calculated
detection limit at the signal-to-noise (S/N) of 3 was 160–370 ppm.
Of course, any accelerometer structure will be sensitive to vibrations if a pre-
scribed data-sampling rate is maintained. In chemical detection, vibration-induced
response is not desirable and will constitute an additional noise source. To cancel
the vibration-induced noise, we sampled the sensor response with ~1 kHz sampling
rate, followed by the electronic averaging of about 1000–2000 data points.
To summarize, our results conclusively demonstrate that this physical trans-
ducer, initially developed as an accelerometer, can be successfully applied as a
chemical sensor without vibration effects. The development and optimization of
MEMS structures often takes more than a decade from their first invention to com-
mercial implementations (Hagleitner et al. 2002). Our work should encourage more
analytical chemistry uses of previously developed MEMS devices for other
demanding applications. Such an approach can be easily adapted for detection of
other chemical as well as biological species. These chemical or biological sensors
also have advantages of simplicity of integration with other types of existing
micromachined sensors to achieve multiparameter measurements and production of
such sensor platforms at low cost.
Radiant Energy Transduction
Optical sensors employ a sensing reagent that undergoes a change in its optical
property (e.g., elastic or inelastic scattering, absorption, luminescence intensity,
luminescence lifetime, or polarization state) upon interaction with the analyte species
(Potyrailo et al. 1998a, 2005d; Potyrailo and Hassib 2005; Potyrailo and Lemmon
2005). Organic reagents are widely employed in existing optical sensors (Bakker
et al. 1997; Wolfbeis 2004), however, such reagents often suffer from their intrinsic
limitations such as rapid photobleaching and short shelf-life. Application of engi-
neered nanostructures provides exciting opportunities for sensor materials with
previously unavailable capabilities (Convertino et al. 2003; Holtz and Asher 1997;
Jiang et al. 2005).
Nanomaterials promise to introduce a variety of new important properties into
sensor response such as improved selectivity, sensitivity, dynamic response, and
long-term stability (Potyrailo 2006). We recently introduced a new concept for
selective chemical sensing based on different size CdSe semiconductor nanocrys-
tals incorporated into a polymer matrix to overcome a photobleaching limitation of
traditional organic reagent-based sensor materials. We have found that when differ-
ent size CdSe nanocrystals made and passivated with tri-n-octylphosphine oxide
(TOPO) using known methods (Kovalevskij et al. 2004; Murray et al. 1993) were
128 R.A. Potyrailo
incorporated into a polymer film and photoactivated, each size of CdSe nanocrys-
tals unexpectedly demonstrated its own photoluminescence (PL) response pattern
upon exposure to polar and nonpolar vapors in air. When this composite response
was processed using multivariate analysis, a single film with different size CdSe
nanocrystals served as a selective sensor.
Our sensing films were produced by spin-casting of a solution of 2.8 and 5.6 nm
diameter nanocrystals and polymethylmethacrylate (PMMA) in toluene and were
photoactivated with a portable 407 nm diode laser. The fabricated sensor films were
exposed to methanol and toluene vapors at the vapor pressure of 46 and 11 torr,
respectively. This corresponded to concentrations of 61,000 and 14,000 ppm of
methanol and toluene, respectively. When a single wavelength was selected at the
peaks of PL spectra of each size of the nanocrystals, two kinetic responses were
obtained. Fig. 6.12a shows baseline-corrected response patterns of chemically
dependent PL of the two-size CdSe nanocrystals sensor film. Emission of the 2.8
nm nanocrystals was measured at 511 nm and emission of the 5.6 nm nanocrystals
was measured at 617 nm. We believe that the difference in the response patterns of
the nanocrystals is related to the combined effects of the dielectric medium sur-
rounding the nanocrystals, their size, and surface oxidation state. The sensitivity of
these sensors was defined as a signal change upon exposure to the known concen-
trations of vapors. The response and recovery kinetics of PL from the 2.8 nm
nanocrystals in PMMA upon exposure to methanol were very fast (<0.5 min).
However, 5.6 nm nanocrystals in the same sensor film exhibited much longer
response and recovery times upon interactions with methanol, 4 and 20 min,
respectively. The 5.6 nm nanocrystals had 4 min response and 0.5 min recovery
times upon interactions with toluene.
Computational studies of the effects of dielectric media on CdSe nanocrystal elec-
tronic properties show that the absolute value of the dipole moment of the nanocrystal
increases with elevated dielectric constant of the surrounding environment (Rabani
et al. 1999). The magnitude of the change in dipole moment increases with particle
size. The elevated diameter and surface area of larger particles means that even small
changes in the surface charge of the nanocrystals will result in significant changes in
the particle’s dipole moment. Both sizes of nanocrystals in our sensor films exhibited
a decrease in emission upon exposure to methanol (e = 33), which causes a large
increase in the local dielectric constant relative to the PMMA medium (e = 2.7–3.0).
Toluene (e = 2.2) has a slightly lower dielectric constant than PMMA and resulted in
an increase in the emission of the larger nanocrystals alone. The different size nano-
crystals also can have different coverage with the capping TOPO ligand (Leatherdale
and Bawendi 2001) and when immobilized into a polymer matrix (Kovalevskij et al.
2004), can contribute to the variable response pattern.
To obtain selective chemical response from individual response patterns of nanoc-
rystals of different size, we applied a principal components analysis (PCA) technique
(Beebe et al. 1998) to analyze the dynamic response data of the sensor film during its
repetitive exposures to methanol and toluene vapors. For PCA analysis, 482 PL spectra
(450–700 nm range) collected during the repetitive (n = 3) exposures of the sensor film
to methanol and toluene vapors were used without baseline drift correction. The capa-
(a) 1500
5.6 nm
1000
PL (counts)
500 2 2 2
1 1 1
2.8 nm
0
0 50 100 150 200
Time (min)
(b) 5
0
PC 3 ( 5.09%)
−5 1
2
−10
−15
−20
−15 −10 −5 0 5 10 15 20
PC 1 (53.96%)
(c) 1400 22
1200
230
1000 11
PL (counts)
220 11
PL (counts)
22
800 210
200
600
190
400 180
0 0.5 1 1.5 2 2.5 3 3.5
Time (h)
200
0
0 2 4 6 8 10 12 14 16
Time (h)
Fig. 6.12 Photoluminescence response of two-size (2.8 and 5.6 nm) CdSe nanocrystals in
poly(methyl methacrylate) polymer film upon repetitive exposures to methanol (1) and toluene (2)
vapors: (a) dynamic vapor response with emission of 2.8 nm nanocrystals at 511 nm and emission
of 5.6 nm nanocrystals at 617 nm. (b) Principal components scores plot demonstrating selective
film response to methanol and toluene. Unlabeled data points result from times when the film was
exposed to a blank (dry air). (c) Long-term stability of film response upon repetitive exposures to
methanol and toluene under a 16 h continuous 407 nm laser excitation
130 R.A. Potyrailo
bility for vapor discrimination using a single sensor film was evaluated from a PCA
model generated with autoscaled data that was built based on the full spectra dynamic
data. The relationship between the collected data was described by plotting scores of
relevant principal components of the PCA model versus each other as shown in Fig.
6.12b. Data analysis with the whole uncorrected spectra provided excellent vapor
discrimination capability due to the elimination of the PL drift-related artifacts. The
remaining response scatter in regions 1 and 2 was due to nonequilibrated sensor
responses during the kinetic experiments. The largest variance (captured by the first
principal component of the PCA model) exhibited the strongest response to both
vapors. The second PC captured the drift contributions in the response.
We further tested the stability of PL emission of new sensor films upon an
extended irradiation with the laser light. Such response stability is critical in continu-
ous monitoring applications. The parameters of interest in these evaluations were:
(1) the overall stability of the PL intensity and (2) the stability of the response pattern
to methanol and toluene vapors. Fig. 6.12c shows response patterns from 2.8 and 5.6
nm nanocrystals in PMMA film over 16 h of continuous exposure of the film to laser
radiation. During the laser exposure, the sensor film was periodically exposed to
methanol and toluene vapors. These data indicated that the response of new selective
sensor films was very stable upon extended exposure to laser excitation without
significant degradation of both emission intensity and response pattern.
Thus, we demonstrated selective detection of polar and nonpolar vapors by gen-
erating highly stable multivariate response patterns from different-size CdSe nanoc-
rystals incorporated in PMMA. Our studies are in progress to better understand the
effects of polymer matrices. Such work promises to complement existing solvato-
chromic organic dye sensors with more photostable and reliable sensor materials.
In our other application of radiant transduction, we have developed an approach
for using unmodified optical disk drives for chemical sensing of water and air
contamination (see Fig. 6.2f). In this sensor system, the analog signals were
acquired from conventional optical disk drives and these signals were used for
quantitative detection of optical changes of sensor films deposited on conventional
CD and DVD optical disks (Potyrailo et al. 2005a,b,c; Potyrailo et al. 2006).
Although the drives still perform their original function of reading and writing
digital content to optical media, they also provide analog signals for quantitative
sensor applications when sensor films are deposited onto optical disks. Because no
alteration of the manufacturing process of optical disks is required, any disk can be
employed for deposition and readout of sensor films. Such a sensor platform is
quite universal and can be applied for quantitative chemical and biological detec-
tion, as well as for the monitoring of changes of physical properties of regions
deposited onto a CD or DVD (e.g., during combinatorial screening of materials).
Fig. 6.13 depicts our concept for chemical and biological detection that employs
an analog signal from a conventional CD/DVD drive to quantify optical changes in
sensor films deposited on the read surface of CD or DVD disks. A conventional
optical disk drive for reading DVDs and CDs contains all needed components to
perform quantitative chemical and biological analysis. The drive has two lasers, 650
and 780 nm to read DVDs and CDs, respectively, a Si photodiode detector, and a
(a) Sensing films
Optical disk Reflective layer
Disc drive motor

Laser pickup head
Focusing optics
Polarizing beam splitter
Laser diode
Tracking drive
Processor Photodetector
EIDE
interface Standard
optical disk drive
Control of
optical disk drive
A/D
LabVIEW
Data acquisition
and drive control Analog
Analog signal acquisition signal
and drive control
(b) Disk Label
Reflective
layer
Sensor film
Detected laser light
(c) Optical phenomena
Amplitude Polarization Phase
Parameters of sensor film

applicable for chemical / biological quantification
in an optical disk drive
x
de
ce
ity
ss
g
in
rin
an
iv
ne
ive
ct
te
rb
ick
le
ct
at
so
ef
Th
fra
Sc
Ab
Re
Fig. 6.13 Concept for quantitative chemical detection using a conventional optical disk drive and
a DVD or CD disk. (a) Schematic of a conventional optical disk drive and the methodology for
obtaining an analog signal from a photodiode detector and for controlling the optical disk drive. (b)
Double-pass interaction of a laser beam with the sensor film deposited onto the read side of an
optical disk. (c) Optical phenomena and parameters of sensor films involved in signal generation
132 R.A. Potyrailo
sophisticated laser tracking system to scan across the disk surface. In the developed
Lab-on-DVD sensor system, an analog signal from the photodiode is extracted
before it is digitized during the reading of the digital content from an optical disk and
brought into a data acquisition program as shown in Fig. 6.13a. This signal is used
for quantitative detection of changes of optical properties of chemical or biological
sensor films deposited on the read surface of the optical disk. In the Lab-on-DVD
sensor system, the optical disk drive is further controlled through the Enhanced
Integrated Disk Electronics (EIDE) interface. The controlled parameters of the opti-
cal disk drive include positioning of the laser pickup head at any specified radial
position, scanning the laser pickup head over a range of desired radii with a control-
led spatial resolution, and the linear rotation velocity of the optical disk.
In the absence of a sensor film, laser light is transmitted through the surface of
the optical disk, reflected from the disk’s reflective data layer and returned to the
photodiode detector. When a sensor film is applied onto the read side of the CD or
DVD disk, the laser light travels through the sensor film twice as shown in Fig.
6.13b. Upon interactions of the sensor film with chemical or biological species,
optical properties of the sensor film vary causing the change in the amount of light
detected by the photodiode detector of the laser pickup head and allowing for quan-
titation of sensor film response.
The optical system of a conventional CD/DVD drive focuses the laser light onto
a reflective layer inside the disk to a spot of about 1 mm and provides polarization
and phase control of the light that reaches the detector. In reading digital data from
a disk, these features are important for the rejection of ambient light and light
produced by scratches and other imperfections on the disk surface. In the current
application of optical disk drives, these features provide an opportunity for the
chemical and biological quantification based on a variety of optical phenomena that
can be produced in sensor films (see Fig. 6.13c).
Numerous possibilities exist for the use of conventional sensor materials as well
as nanomaterials for the realization of these optical phenomena. Color changes can
be produced by a wide variety of organic dyes developed in the twentieth century for
wet chemistry test methods and adapted for solid-film sensing (Capitán-Vallvey
et al. 2000; Chau and Porter 1990; Dybko et al. 1998). Plasmon resonance bands of
metal nanoparticles and other plasmonic nanostructures can be tailored to 650 and
780 nm laser wavelengths and band shifts can be produced by the aggregation or
deaggregation of these plasmonic nanostructures in sensing films (Elghanian et al.
1997; Nath and Chilkoti 2002). Photonic crystals can be easily assembled with their
diffraction peaks in the vicinity of the laser wavelengths with biochemical reactions
inducing peak shifts toward or from the laser wavelengths (Holtz and Asher 1997).
Variations in other optical phenomena such as scattering produced by phase and
morphology change in polymers (Potyrailo et al. 2002), metal film reflectivity
produced by film degradation (Bouten et al. 2002) or silver staining (Taton et al.
2000), and thickness and refractive index produced by polymer swelling (McCurley
and Seitz 1991; Miyata et al. 1999; Zhang et al. 2003) have much smaller wave-
length dependence and can be applied with any laser in the optical disk drive. Newly
introduced Blu-Ray® optical disk drives with their 405 nm lasers are also attractive
in expanding the range of optical sensing materials (e.g., freebase and metallo
porphyrins) for chemical and biological sensing using new-generation optical disks.
It is worth noting that we do not measure fluorescence or other types of emission
from the sensor films using the unmodified optical disk drives due to the optical
properties of the disk drive. However, certain types of new nanoparticle-based
colorimetric and other principles that can provide sensitivity comparable to or
greater than fluorescence can be employed (Steinberg et al. 1996; Taton et al. 2000;
Wetzl et al. 2004).
Performance of the Lab-on-DVD sensor system has been demonstrated for
quantitative chemical analysis in water and air. For example, for determination of
Ca2+, sensing films incorporated a Ca2+ sensitive dye Xylidyl Blue in a poly-
(2-hydroxyethy) methacrylate hydrogel matrix. For fabrication of films, the dye/
polymer/solvent (1-methoxy-2-propanol) solution was applied to the DVDs to
form 3 × 4 mm sensing films. After solvent evaporation, the sensing films
adhered to the DVD surface and were ready for testing for their response to Ca2+.
Water samples (20–40 mL volumes) with varied concentrations of Ca2+ were
applied to the sensing films and removed with pressurized house nitrogen gas
after 2 min of exposure (Potyrailo et al. 2006). After exposure, quantitative meas-
urements were performed in the optical disk drive. A typical calibration curve for
Ca2+ determinations is presented in Fig. 6.14a. The shape of the calibration curve
was typical of determinations of Ca2+ and other cations using organic chromog-
enic dyes immobilized in polymeric films (Seitz 1988; Wolfbeis 1991). The cal-
culated detection limit for Ca2+ determinations (at S/N = 3) was 5 ppm for the
Lab-on-DVD measurement system. This detection limit corresponded to a 0.023
absorbance unit resolution as obtained by measurement of the same colorimetric
films with the reference optical spectrometer. This detection limit in Ca2+ deter-
minations is applicable for both clinical and environmental applications.
In another application of Lab-on-DVD, for the determination of chlorine in
water, we have developed a indicator formulation that contained 1¢,1¢-diethyl-
4,4¢-carbocyanine iodide dye in a poly(2-hydroxyethymethacrylate) hydrogel
matrix. The indicator concentration in the film was optimized for detection of
chlorine at ppb concentrations. A calibration curve for chlorine determinations
with the Lab-on-DVD system is presented in Fig. 6.14b. Indicator film responses
were recorded after 2 min exposures to chlorine-containing water solutions (~30
mL volumes). Each data point was a result of measurements performed at three
radial positions separated by 0.5 mm. Typical noise in the determinations was
<10% RSD. The calculated detection limit for chlorine determinations (at S/N =
3) was 200 ppb.
We have also developed an approach for measurements of chemical species in
air using Lab-on-DVD. In this application, a vapor-introduction port has been made
into an optical disk drive and different concentrations of tested vapors were intro-
duced into the optical disk drive. A demonstration of this sensing approach has
been performed with quantitation of water vapor in air. A sensing film comprised
of a Nafion polymer formulated with rhodamine 800 dye has been applied to a
DVD. A signal change from the detector has been recorded in real-time upon
134 R.A. Potyrailo
(a)
100
Normalized DVD Signal

80
60
40
20
0
1 10 100 1000
Calcium Conc. (ppm)
(b)
100
75
50
25
0 200 400 600 800 1000

Chlorine Conc. (ppb)
76 (c)
72
68
0 10 20 30 40 50 60 70
64 % RH
60
56 1 2 1 2
52
0 200 400 600 800 1000 1200
Time (s)
Fig. 6.14 Quantitation of chemical species in water and air using Lab-on-DVD sensor system. (a)
Ca2+ determinations in water. (b) Chlorine determinations in water. (c) Relative humidity determina-
tions in air. Each data point in (a) and (b) is the mean of three replicate measurements across different
radial positions of an indicator film; error bars are one standard deviation. Numbers in (c) are relative
humidity values: (1) 22% RH and (2) 67% RH. Baseline is response of the sensing film in dry air
exposure of the DVD with the sensing film to different relative humidity while the
DVD was reading in the optical disk drive. A reversible signal change has been
observed of the absorption of the sensing film as shown in Fig. 6.14c with the signal
change proportional to the relative humidity of air.
In conclusion, Lab-on-DVD technology coupled with diverse indicator film chem-

istries offers several practical advantages as compared to use of the same sensor films
with other common detection platforms such as plate readers and flatbed scanners.
Summary
Although this review summarized recent achievements in the area of chemical

sensors in our laboratory, these developments follow more general trends in the
modern design of chemical sensors.
In the area of sensing materials, nanofabricated, nanocomposite, and nanoassem-
bled sensing materials promise to provide previously unavailable sensing capabili-
ties. For example, an enhanced photostability is expected from colloidal crystal
array chemically sensing films because the colorimetric response in these films is
generated due to the physical effects rather than due to the chemical changes in
organic dyes (Potyrailo et al. 2007a). Also, new opportunities for highly selective
vapor detection are becoming available with carefully designed nanostructured
vapor-responding films (Potyrailo et al. 2007b). Furthermore, unexpected new
properties of sensing materials can be achieved from small-diameter electrospun
composite nanofibers (Potyrailo 2006). Although rational design of sensing materials
remains a big challenge, occasional successes have been reported (Potyrailo 2006).
Rational design of sensing materials will continue to improve with the development
of new molecular modeling and data mining tools.
In the area of physical transducers, advantages of miniaturization of transducers
will be more practiced and limitations of miniaturization of transducers will be
more understood. The advances in integrated-circuit technologies, in particular,
improved precision and quality of fabrication with nanometer resolution as well as
manipulation of minute objects on a nanoscale will be further translated into sens-
ing applications. However, the compatibility of established integrated-circuit tech-
nologies (e.g., based on Si) will continue to be explored for sensing applications.
New knowledge will be established that should provide further understanding of the
interface bonds between sensing and substrate materials (Göpel 1996). Of course,
the ultimate goal is to develop fully integrated sensor systems that have integrated
system components that perform sensing, data processing, and actuation (Göpel
1996; Hagleitner et al. 2001). However, from the transducer design perspective, it
is also critical to keep in mind that the sensitivity is not the most important aspect
for quantitative measurements with micro- and nanofabricated transducers, but
rather their signal-to-noise ratio. Thus, care should be taken while designing the
nanofabricated transducers because some transducers are scaling invariant, whereas
others are scaling sensitive. For example, electrical energy transducers such as
potentiometric devices that measure voltage (ion-selective electrodes, ion-sensitive
field effect transistors, etc.) are scaling invariant, and amperometric devices that
measure a current, are scaling sensitive. In radiant transducers, scaling invariants are
those that measure the refractive index whereas many others (emission, absorption, etc.)
are scaling sensitive (Madou 2002; Madou and Cubicciotti 2003).
136 R.A. Potyrailo
New analyte sampling concepts will be required to address transducer miniaturi-

zation and the need for analyte detection at ultratrace levels with enhanced selectivity
and better signal drift-free performance. New models have been developed for the
detection of large analyte molecules in solutions at ultratrace concentrations
(Franke et al. 2006; Hassibi et al. 2004; Sheehan and Whitman 2005; Vikalo et al.
2006; Zahedi et al. 2004). New models will be further developed to guide the
design of future sensor systems to ultratrace chemical detection. Typically, with the
detection of small concentrations, the signal-to-noise ratio of these determinations
decreases. Thus, the sensor drift becomes more noticeable. New approaches will be
needed to overcome or reduce drift effects. One of the promising approaches is to
use a time-modulated sample introduction with the periodic establishment of a
baseline in sensor response (Grate et al. 1993; Groves et al. 1998; Kindlund et al.
1984; Potyrailo and May 2002; Potyrailo and Sivavec 2005).
Finally, in the data handling area, the need will be to develop new concepts to extract
more selective quantitative information from the sensors’ systems when quantifying
multiple analytes in their mixtures and in the presence of interferences. Today, using a
sensor array with partially selective sensing films and a single transduction principle,
one can expect to reliably quantify four or fewer analytes in their mixtures (Grate 2000;
Hsieh and Zellers 2004; Park et al. 1999). Thus, higher-order sensor systems (Booksh
and Kowalski 1994) will continue to attract attention where additional dimensions are
added to the discrimination response through hyphenated transduction techniques and
time-, temperature-, sample-flow, and other signal modulation approaches.
Acknowledgments This research has been inspired by the creative teammates at GE Global
Research, Nomadics, and Indiana University, Bloomington, IN who have coauthored original
contributions cited here: S. Boyette, M. D. Butts, J. R. Cournoyer, Z. Ding, K. Dovidenko,
W. P. Flanagan, S. K. Gamage, S. E. Genovese, L. Hassib, A. M. Leach, J. P. Lemmon, R. J. May,
W. G. Morris, E. Olson, J. J. Salvo, O. P. Siclovan, R. E. Shaffer, T. M. Sivavec, A. Vertiatchikh,
M. B. Wisnudel, and R. J. Wroczynski (GE), L. Salsman (Nomadics), R. C. Conrad, T. L.
Danielson, M. Johnson, and A. W. Szumlas (Indiana University), and H. Ghiradella (SUNY
Albany). Special thanks go to G. M. Hieftje (Indiana University), A. D. Ellington (while at Indiana
University),T. K. Leib, and A. Linsebigler (GE) for letting the creativity grow and expand.
References
Adhikari B, Majumdar S (2004) Polymers in sensor applications. Prog. Polym. Sci. 29:699–766.
Ahuja A, James DL, Narayan R (1999) Dynamic behavior of ultra-thin polymer films deposited
on surface acoustical wave devices. Sens. Actuators B 72:234–241.
Archibald B, Brümmer O, Devenney M, Giaquinta DM, Jandeleit B, Weinberg WH, Weskamp T
(2002a) Combinatorial aspects of materials science. In: Nicolaou KC, Hanko R, Hartwig W,
editors. Handbook of Combinatorial Chemistry. Drugs, Catalysts, Materials. Weinheim: Wiley,
chapter 34, pp. 1017–1062.
Archibald B, Brümmer O, Devenney M, Gorer S, Jandeleit B, Uno T, Weinberg WH, Weskamp T
(2002b) Combinatorial methods in catalysis. In: Nicolaou KC, Hanko R, Hartwig W, editors.
Handbook of Combinatorial Chemistry. Drugs, Catalysts, Materials. Weinheim: Wiley, chapter
32, pp. 885–990.
Artmann R (1999) Electronic identification systems: State of the art and their further development.
Comput. Electron. Agri. 24:5–26.
Bakker E, Bühlmann P, Pretsch E (1997) Carrier-based ion-selective electrodes and bulk optodes.
1. General characteristics. Chem. Rev. 97:3083–3132.
Ballantine DS, Jr., White RM, Martin SJ, Ricco AJ, Frye GC, Zellers ET, Wohltjen H (1997)
Acoustic Wave Sensors: Theory, Design, and Physico-chemical Applications. San Diego, CA:
Academic Press. 436 p.
Beebe KR, Pell RJ, Seasholtz MB (1998) Chemometrics: A Practical Guide. New York: Wiley.
Bender F, Barié N, Romoudis G, Voigt A, Rapp M (2003) Development of a preconcentration
unit for a SAW sensor micro array and its use for indoor air quality monitoring. Sens.
Actuators B 76:1–7.
Bennett MD, Leo DJ (2004) Ionic liquids as stable solvents for ionic polymer transducers. Sens.
Actuators A 115:79–90.
Benschop J, Rosmalen GV (1991) Confocal compact scanning optical microscope based on
compact disc technology. Appl. Opt. 30:1179–1184.
Booksh KS, Kowalski BR (1994) Theory of analytical chemistry. Anal. Chem. 66:782A–791A.
Boussaad S, Tao NJ (2003) Polymer wire chemical sensor using a microfabricated tuning fork.
Nano Lett. 3:1173–1176.
Capitán-Vallvey LF, de Cienfuegos-Gálvez PA, Fernández Ramos MD, Avidad-Castañeda R (2000)
Determination of calcium by a single-use optical sensor. Sens. Actuators B 71:140–146.
Carrano JC, Jeys T, Cousins D, Eversole J, Gillespie J, Healy D, Licata N, Loerop W, O’Keefe M,
Samuels A, Schultz J, Walter M, Wong N, Billotte B, Munley M, Reich E, Roos J (2004)
Chemical and Biological Sensor Standards Study (CBS3). In: Carrano JC, Zukauskas A, edi-
tors. Optically Based Biological and Chemical Sensing for Defence. Bellingham, WA: SPIE
– The International Society for Optical Engineering. pp. xi–xiii.
Chang TC, Wang GP, Hong YS, Chen HB (2001) Characterization and degradation of hydrogen-
bonded acidic polyamideimides linked by disiloxanes. Polym. Degrad. Stab. 73:301–308.
Chau LK, Porter MD (1990) Optical sensor for calcium: Performance, structure, and reactivity of
calcichrome immobilized at an anionic polymer film. Anal. Chem. 62:1964–1971.
Cho EJ, Bright FV (2001) Optical sensor array and integrated light source. Anal. Chem.
73:3289–3293.
Cho EJ, Tao Z, Tehan EC, Bright FV (2002) Multianalyte pin-printed biosensor arrays based on
protein-doped xerogels. Anal. Chem. 74:6177–6184.
Christensen CM (1997) The Innovator’s Dilemma: When New Technologies Cause Great Firms
to Fail. Boston, MA: Harvard Business School Press.
Christensen CM, Raynor ME (2003) The Innovator’s Solution: Creating and Sustaining Successful
Growth. Boston, MA: Harvard Business School Press.
Chu C-L, Lin C-H (2005) Development of an optical accelerometer with a DVD pick-up head.
Meas. Sci. Technol. 16:2498–2502.
Convertino A, Capobianchi A, Valentini A, Cirillo ENM (2003) A new approach to organic sol-
vent detection: High-reflectivity Bragg reflectors based on a gold nanoparticle/Teflon-like
composite material. Adv. Mater. 15:1103–1105.
DeLongchamp DM, Hammond PT (2003) Fast ion conduction in layer-by-layer polymer films.
Chem. Mater. 15:1165–1173.
Dutta P, Tipple CA, Lavrik NV, Datskos PG, Hofstetter H, Hofstetter O, Sepaniak MJ (2003)
Enantioselective sensors based on antibody-mediated nanomechanics. Anal. Chem.
75:2342–2348.
Dybko A, Wróblewski W, Rozniecka E, Pozniakb K, Maciejewski J, Romaniuk R, Brzózka Z
(1998) Assessment of water quality based on multiparameter fiber optic probe. Sens. Actuators
B 51:208–213.
Elghanian R, Storhoff JJ, Mucic RC, Letsinger RL, Mirkin CA (1997) Selective colorimetric
detection of polynucleotides based on the distance-dependent optical properties of gold nano-
particles. Science 277:1078–1081.
Erickson D, Li D (2004) Integrated microfluidic devices. Anal. Chim. Acta 507:11–26.
138 R.A. Potyrailo
Fatibello-Filho O, de Andrade JF, Suleiman AA, Guilbault GG (1989) Piezoelectric crystal moni-
tor for carbon dioxide in fermentation processes. Anal. Chem. 61:746–748.
Feng C-D, Sun S-L, Wang H, Segre CU, Stetter JR (1997) Humidity sensing properties of nation
and sol-gel derived SiO2/Nafion composite thin films. Sens. Actuators B 40:217–222.
Filippini D, Alimelli A, Di Natale C, Paolesse R, D’Amico A, Lundström I (2006) Chemical sens-
ing with familiar devices. Angew. Chem. Int. Ed. 45:3800–3803.
Finkenzeller K (2003) RFID Handbook. Fundamentals and Applications in Contactless Smart
Cards and Identification. Hoboken, NJ: Wiley.
Finklea HO, Phillippi MA, Lompert E, Grate JW (1998) Highly sorbent films derived from
Ni(SCN)2(4-picoline)4 for the detection of chlorinated and aromatic hydrocarbons with quartz
crystal microbalance sensors. Anal. Chem. 70:1268–1276.
Franke ME, Koplin TJ, Simon U (2006) Metal and metal oxide nanoparticles in chemiresistors:
Does the nanoscale matter? Small 2:36–50.
Frantzen A, Scheidtmann J, Frenzer G, Maier WF, Jockel J, Brinz T, Sanders D, Simon U (2004)
High-throughput method for the impedance spectroscopic characterization of resistive gas
sensors. Angew. Chem. Int. Ed. 43:752–754.
Furuki M, Pu LS (1992) Hybrid gas detector of squarylium dye Langmuir-Blodgett film deposited
on a quartz oscillator. Thin Solid Films 210/211:471–473.
Gomes MTSR, Rocha TA, Duarte AC, Oliveira JABP (1996) The performance of a tetramethyl-
ammonium fluoride tetrahydrate coated piezoelectric crystal for carbon dioxide detection.
Anal. Chim. Acta 335:235–238.
Göpel W (1996) Ultimate limits in the miniaturization of chemical sensors. Sens. Actuators A
56:83–102.
Grate JW (2000) Acoustic wave microsensor arrays for vapor sensing. Chem. Rev. 100:2627–2648.
Grate JW, Rose-Pehrsson SL, Venezky DL, Klusty M, Wohltjen H (1993) Smart sensor system for
trace organophosphorus and organosulfur vapor detection employing a temperature-controlled
array of surface acoustic wave sensors, automated sample preconcentration, and pattern recog-
nition. Anal. Chem. 65:1868–1881.
Groves WAGrate JW, Abraham H, McGill RA (1997a) Sorbent polymer materials for chemical
sensors and arrays. In: Kress-Rogers E, editor. Handbook of Biosensors and Electronic Noses.
Medicine, Food, and the Environment. Boca Raton, FL: CRC. pp. 593–612.
Grate JW, Kaganove SN, Bhethanabotla VR (1997b) Examination of mass and modulus contribu-
tions to thickness shear mode and surface acoustic wave vapour sensor responses using parti-
tion coefficients. Faraday Discuss. 107:259–283.
Grate JW, Kaganove SN, Bhethanabotla VR (1998) Comparisons of polymer/gas partition coef-
ficients calculated from responses of thickness shear mode and surface acoustic wave vapor
sensors. Anal. Chem. 70:199–203.
Groves WA, Zellers ET, Frye GC (1998) Analyzing organic vapors in exhaled breath using a
surface acoustic wave sensor array with preconcentration: Selection and characterization of the
preconcentrator adsorbent. Anal. Chim. Acta 371:131–143.
Hagleitner C, Hierlemann A, Lange D, Kummer A, Kerness N, Brand O, Baltes H (2001) Smart
single-chip gas sensor microsystem. Nature 414:293–296.
Hansen CL, Skordalakes E, Berger JM, Quake SR (2002) A robust and scalable microfluidic
metering method that allows protein crystal growth by free interface diffusion. Proc. Natl.
Acad. Sci. USA 99:16531–16536.
Hansen KM, Ji H-F, Wu G, Datar R, Cote R, Majumdar A, Thundat T (2001) Cantilever-based
optical deflection assay for discrimination of DNA single-nucleotide mismatches. Anal.
Chem. 73:1567–1571.
Harsanyi G (1995) Polymer Films in Sensor Applications. Lancaster, PA: Technomic.
Hierlemann A, Baltes H (2003) CMOS-based chemical microsensors. Analyst 128:15–28.
Hirayama E, Sugiyama T, Hisamoto H, Suzuki K (2000) Visual and colorimetric lithium ion sens-
ing based on digital color analysis. Anal. Chem. 72:465–474.
Hirschfeld T (1985) Instrumentation in the next decade. Science 230:286–291.
Hirschfeld T, Callis JB, Kowalski BR (1984) Chemical sensing in process analysis. Science
226:312–318.
Hofmann MC, Kensy F, Buechs J, Mokwa W, Schnakenberg U (2005) Transponder-based sensor for
monitoring electrical properties of biological cell solutions. J. Biosci. Bioeng. 100:172–177.
Holtz JH, Asher SA (1997) Polymerized colloidal crystal hydrogel films as intelligent chemical
sensing materials. Nature 389:829–832.
Hoyt AE, Ricco AJ, Bartholomew JW, Osbourn GC (1998) SAW sensors for the room-tempera-
ture measurement of CO2 and relative humidity. Anal. Chem. 70:2137–2145.
Hruschka WR, Massie DR, Anderson JD (1983) Computerized analysis of two-dimensional elec-
trophoretograms. Anal. Chem. 55:2345–2348.
Hsieh M-D, Zellers ET (2004) Limits of recognition for simple vapor mixtures determined with a
microsensor array. Anal. Chem. 76:1885–1895.
Hughes RC, Yelton WG, Pfeifer KB, Patel SV (2001) Characteristics and mechanisms in ion-
conducting polymer films as chemical sensors. Polyethyleneoxide J. Electrochem. Soc.
148(4):H37–H44.
Ivanisevic A, Yeh J-Y, Mawst L, Kuech TF, Ellis AB (2001) Light-emitting diodes as chemical
sensors. Nature 409:476.
Janata J (1989) Principles of Chemical Sensors. New York: Plenum.
Janata J, Josowicz M (2002) Conducting polymers in electronic chemical sensors. Nat. Mater.
2:19–24.
Jandeleit B, Schaefer DJ, Powers TS, Turner HW, Weinberg WH (1999) Combinatorial materials
science and catalysis. Angew. Chem. Int. Ed. 38:2494–2532.
Jarrett MR, Finklea HO (1999) Detection of nonpolar vapors on quartz crystal microbalances with
Ni(SCN)2(4-picoline)4 coatings. Anal. Chem. 71:353–357.
Jiang P, Smith DWJ, Ballato JM, Foujger SH (2005) Multicolor pattern generation in photonic
bandgap composites. Adv. Mater. 17:179–184.
Jones CL (2005) Cryptographic hash functions and CD-based optical biosensors. Prob. Nonlinear
Anal. Eng. Syst. 2(23):17–36.
Jurs PC, Bakken GA, McClelland HE (2000) Computational methods for the analysis of chemical
sensor array data from volatile analytes. Chem. Rev. 100:2649–2678.
Kindlund A, Sundgren H, Lundstrom I (1984) Quartz crystal gas monitor with gas concentrating
stage. Sens. Actuators 6:1–17.
King WH, Jr. (1964) Piezoelectric sorption detector. Anal. Chem. 36:1735–1739.
Koinuma H, Takeuchi I (2004) Combinatorial solid state chemistry of inorganic materials. Nat.
Mater. 3:429–438.
Korsah K, Ma CL, Dress B (1998) Harmonic frequency analysis of SAW resonator chemical sen-
sors: Application to the detection of carbon dioxide and humidity. Sens. Actuators B
50:110–116.
Kovalevskij V, Gulbinas V, Piskarskas A, Hines MA, Scholes GD (2004) Surface passivation in
CdSe nanocrystal-polymer films revealed by ultrafast excitation relaxation dynamics. Phys.
Stat. Sol. B 241(8):1986–1993.
Kuban P, Berg JM, Dasgupta PK (2004) Durable microfabricated high-speed humidity sensors.
Anal. Chem. 76:2561–2567.
La Clair JJ, Burkart MD (2003) Molecular screening on a compact disc. Org. Biomol. Chem.
1:3244–3249.
Lange SA, Roth G, Wittemann S, Lacoste T, Vetter A, Grässle J, Kopta S, Kolleck M, Breitinger
B, Wick M, Hörber JKH, Dübel S, Bernard A (2006) Measuring biomolecular binding events
with a compact disc player device. Angew. Chem. Int. Ed. 45:270–273.
Lavrik NV, Sepaniak MJ, Datskos PG (2004) Cantilever transducers as a platform for chemical
and biological sensors. Rev. Sci. Instrum. 75:2229–2253.
Leatherdale CA, Bawendi MG (2001) Observation of solvatochromism in CdSe colloidal quantum
dots. Phys. Rev. B 63:165315 1–6.
Mackay RS, Jaconson B (1957) Endoradiosonde. Nature 179:1239–1240.
140 R.A. Potyrailo
Madaras MB, Buck RP (1996) Miniaturized biosensors employing electropolymerized permselec-

tive films and their use for creatinine assays in human serum. Anal. Chem. 68:3832–3839.
Madou MJ (2002) Fundamentals of Microfabrication. The Science of Miniaturization. Boca
Raton, FL: CRC.
Madou MJ, Cubicciotti R (2003) Scaling issues in chemical and biological sensors. Proc. IEEE
91:830–838.
Manzano J, Filippini D, Lundström I (2003) Computer screen illumination for the characterization
of colorimetric assays. Sens. Actuators B 96:173–179.
Mascaro DJ, Baxter JC, Halvorsen A, White K, Scholz B, Schulz DL (2007) ChemiBlock trans-
ducers. Sens. Actuators B 120:353–361.
McCurley MF, Seitz WR (1991) Fiber-optic sensor for salt concentration based on polymer swell-
ing coupled to optical displacement. Anal. Chim. Acta 249:373–380.
McGrath JE, Dunson DL, Mecham SJ, Hedrick JL (1999) Synthesis and characterization of seg-
mented polyimide-polyorganosiloxane copolymers. Adv. Polym. Sci. 140:61–105.
McQuade DT, Pullen AE, Swager TM (2000) Conjugated polymer-based chemical sensors.
Chem. Rev. 100: 2537–2574.
Meyerhoff ME (1993) In vivo blood-gas and electrolyte sensors: Progress and challenges. Trends
Anal. Chem. 12:257–266.
Mitchell GL (Future Focus Inc., Woodinville, WA, USA) Source: Proceedings of the SPIE - The
International Society for Optical Engineering, v 2574, 1995, p 132-40
Middelhoek S, Noorlag JW (1981/1982) Three-dimensional representation of input and output
transducers. Sens. Actuators 2:29–41.
Miyata T, Asami N, Uragami T (1999) A reversibly antigen-responsive hydrogel. Nature
399:766–769.
Morris DR, Sun X (1993) Water-sorption and transport properties of Nafion 117H. J. Appl.
Polym. Sci. 50:1445–1452.
Murray CB, Norris DJ, Bawendi MG (1993) Synthesis and characterization of nearly monodis-
perse CdE (E = sulfur, selenium, tellurium) semiconductor nanocrystallites. J. Am. Chem. Soc.
115:8706–8715.
Nambi S, Nyalamadugu S, Wentworth SMm Chin BA (2003) Radio frequency identification sen-
sors. Proceedings & Informatics, Orlando, Florida, USA, pp.386–390.
Nath N, Chilkoti A (2002) A colorimetric gold nanoparticle sensor to interrogate biomolecular
interactions in real time on a surface. Anal. Chem. 74:504–509.
Nicolaou KC, Xiao X-Y, Parandoosh Z, Senyei A, Nova MP (1995) Radiofrequency encoded
combinatorial chemistry. Angew. Chem. Int. Ed. 34:2289–2291.
Ong KG, Wang J, Singh RS, Bachas LG, Grimes CA (2001) Monitoring of bacteria growth using
a wireless, remote query resonant-circuit sensor: Application to environmental sensing.
Biosens. Bioelectron. 16:305–312.
Opekar F, Štulik K (1999) Electrochemical sensors with solid polymer electrolytes. Anal. Chim.
Acta 385:151–162.
Oprea A, Henkel K, Oehmgen R, Appel G, Schmeißer D, Lauer H, Hausmann P (1999) Increased
sensor sensitivities obtained by polymer-coated quartz microbalances. Mat. Sci. Eng. C
8–9:509–512.
Park J, Groves WA, Zellers ET (1999) Vapor recognition with small arrays of polymer-coated
microsensors. A comprehensive analysis. Anal. Chem. 71:3877–3886.
Park J, Hsu T, Groves WA (1998) Establishing a limit of recognition for a vapor sensor array.
Anal. Chem. 70:4191–4201.
Patrash SJ, Zellers ET (1993) Characterization of polymeric surface acoustic wave sensor coatings
and semiempirical models of sensor responses to organic vapors. Anal. Chem. 65:2055–2066.
Peterson DS (2005) Solid supports for micro analytical systems. Lab Chip 5:132–139.
Pickup JC, Alcock S (1991) Clinicians’ requirements for chemical sensors for in vivo monitoring:
A multinational survey. Biosens. Bioelectron. 6:639–646.
Piner RD, Zhu J, Xu F, Hong S, Mirkin CA (1999) Dip-pen nanolithography. Science
283:661–663.
Potyrailo RA (2006) Polymeric sensor materials: Toward an alliance of combinatorial and rational
design tools? Angew. Chem. Int. Ed. 45:702–723.
Potyrailo RA, Hassib L (2005) Analytical instrumentation infrastructure for combinatorial and
high-throughput development of formulated discrete and gradient polymeric sensor materials
arrays. Rev. Sci. Instrum. 76:062225.
Potyrailo RA, Lemmon JP (2005) Time-modulated combinatorially developed optical sensors for
determination of non-volatile analytes in complex samples. QSAR Comb. Sci. 24:7–17.
Potyrailo RA, May RJ (2002) Dynamic high throughput screening of chemical libraries using
acoustic-wave sensor system. Rev. Sci. Instrum. 73:1277–1283.
Potyrailo RA, Morris WG (2007) Multianalyte chemical identification and quantitation using a
single radio frequency identification sensor. Anal. Chem. 79:45–51.
Potyrailo RA, Sivavec TM (2004) Boosting sensitivity of organic vapor detection with silicone
block polyimide polymers. Anal. Chem. 76:7023–7027.
Potyrailo RA, Sivavec TM (2005) Dual-response resonant chemical sensors for multianalyte
analysis. Sens. Actuators B 106:249–252.
Potyrailo RA, Takeuchi I, editors. (2005) Special feature on combinatorial and high-throughput
materials research. Meas. Sci. Technol. 316: 1–4
Potyrailo RA, Chisholm BJ, Olson DR, Brennan MJ, Molaison CA (2002) Development of com-
binatorial chemistry methods for coatings: High-throughput screening of abrasion resistance
of coatings libraries. Anal. Chem. 74:5105–5111.
Potyrailo RA, Conrad RC, Ellington AD, Hieftje GM (1998a) Adapting selected nucleic acid
ligands (aptamers) to biosensors. Anal. Chem. 70:3419–3425.
Potyrailo RA, Ding Z, Butts MD, Genovese SE (2008) Selective chemical sensing using structur-
ally colored core-shell colloidal crystal films, 8:815–22.
Potyrailo RA, Ghiradella H, Vertiatchikh A, Dovidenko K, Cournoyer JR, Olson E (2007b)
Morpho butterfly wing scales demonstrate highly selective vapour response. Nat. Photon.
1:123–128 (Cover Story).
Potyrailo RA, Hobbs SE, Hieftje GM (1998b) Optical waveguide sensors in analytical chemistry:
Today’s instrumentation, applications and future development trends. Fresenius’ J. Anal.
Chem. 362:349–373.
Potyrailo RA, Leach AM, Morris WG, Gamage SK (2006a) Chemical sensors based on
micromachined transducers with integrated piezoresistive readout. Anal. Chem. 78:
5633–5638.
Potyrailo RA, May RJ, Sivavec TM (2004b) Recognition and quantification of perchloroethylene,
trichloroethylene, vinyl chloride, and three isomers of dichloroethylene using acoustic-wave
sensor array. Sens. Lett. 2:31–36.
Potyrailo RA, Morris WG, Leach AM, Sivavec TM, Wisnudel MB, Boyette S (2006b) Analog
signal acquisition from computer optical drives for quantitative chemical sensing. Anal. Chem.
78:5893–5899.
Potyrailo RA, Morris WG, Wroczynski RJ (2004c) Multifunctional sensor system for high-
throughput primary, secondary, and tertiary screening of combinatorially developed materials.
Rev. Sci. Instrum. 75:2177–2186.
Potyrailo RA, Szumlas AW, Danielson TL, Johnson M, Hieftje GM (2005d) A dual-parameter
optical sensor fabricated by gradient axial doping of an optical fibre. Meas. Sci. Technol.
16:235–241.
Quercioli F, Tiribilli B, Ascoli C, Baschieri P, Frediani C (1999) Monitoring of an atomic force
microscope cantilever with a compact disk pickup. Rev. Sci. Instrum. 70:3620–3624.
Rabani E, Hetényi B, Berne BJ, Brus LE (1999) Electronic properties of CdSe nanocrystals in the
absence and presence of a dielectric medium. J. Chem. Phys. 110:5355–5369.
Rakow NA, Suslick KS (2000) A colorimetric sensor array for odour visualization. Nature
406:710–713.
Ren M, Forzani ES, Tao N (2005) Chemical sensor based on microfabricated wristwatch tuning
forks. Anal. Chem. 77:2700–2707.
142 R.A. Potyrailo
Rosler S, Lucklum R, Borngraber R, Hartmann J, Hauptmann P (1998) Sensor system for the
detection of organic pollutants in water by thickness shear mode resonators. Sens. Actuators
B 48:415–424.
Savran CA, Knudsen SM, Ellington AD, Manalis SR (2004) Micromechanical detection of pro-
teins using aptamer-based receptor molecules. Anal. Chem. 76:3194–3198.
Scheidtmann J, Frantzen A, Frenzer G, Maier WF (2005) A combinatorial technique for the search
of solid state gas sensor materials. Meas. Sci. Technol. 16:119–127.
Seitz WR (1988) Chemical sensors based on immobilized indicators and fiber optics. CRC Crit.
Rev. Anal. Chem. 19(2):135–173.
Sepaniak MJ, Datskos PG, Lavrik NV, Tipple C (2002) Microcantilever transducers: A new
approach in sensor technology. Anal. Chem. 74:568A–575A.
Sheehan PE, Whitman LJ (2005) Detection limits for nanoscale biosensors. Nano Lett.
5:803–807.
Singh K, Shahi VK (1998) Electrochemical studies on nation membrane. J. Memb. Sci.
140:51–56.
Smit MA, Ocampo AL, Espinosa-Medina MA, Sebastián PJ (2003) A modified Nafion membrane
with in situ polymerized polypyrrole for the direct methanol fuel cell. J. Power Sources
124:59–64.
Steinberg TH, Jones LJ, Haugland RP, Singer VL (1996) SYPRO orange and SYPRO red protein
gel stains: One-step fluorescent staining of denaturing gels for detection of nanogram levels of
protein. Anal. Biochem. 239:223–237.
Su M, Li S, Dravid VP (2003) Microcantilever resonance-based DNA detection with nanoparticle
probes. Appl. Phys. Lett. 82:3562–3564.
Su P-G, Sun Y-L, Lin C-C (2006) A low humidity sensor made of quartz crystal microbalance
coated with multi-walled carbon nanotubes/Nafion composite material films. Sens. Actuators
B 115:338–343.
Suzuki H (2000) Advances in the microfabrication of electrochemical sensors and systems.
Electroanalysis 12:703–715.
Suzuki K, Hirayama E, Sugiyama T, Yasuda K, Okabe H, Citterio D (2002) Ionophore-based
lithium ion film optode realizing multiple color variations utilizing digital color analysis. Anal.
Chem. 74:5766–5773.
Tailoka F, Fray DJ, Kumar RV (2003) Application of Nafion electrolytes for the detection of
humidity in a corrosive atmosphere. Solid State Ionics 161:267–277.
Tani H, Maehana K, Kamidate T (2004) Chip-based bioassay using bacterial sensor strains immo-
bilized in three-dimensional microfluidic network. Anal. Chem. 76:6693–6697.
Taton TA, Mirkin CA, Letsinger RL (2000) Scanometric DNA array detection with nanoparticle
probes. Science 289:1757–1760.
Thompson M, Stone DC (1997) Surface-Launched Acoustic Wave Sensors: Chemical Sensing and
Thin-Film Characterization. New York: Wiley. 196 p.
Thorsen T, Maerkl SJ, Quake SR (2002) Microfluidic large-scale integration. Science
298:580–584.
Tricoli V, Nannetti F (2003) Zeolite-Nafion composites as ion conducting membrane materials.
Electrochim. Acta 48:2625–2633.
Vikalo H, Hassibi B, Hassibi A (2006) A statistical model for microarrays, optimal estimation
algorithms, and limits of performance. IEEE Trans. Signal Proc. 54:2444–2455.
Vilkner T, Janasek D, Manz A (2004) Micro total analysis systems. Recent developments. Anal.
Chem. 76:3373–3386.
Vo-Dinh T, Alarie JP, Isola N, Landis D, Wintenberg AL, Ericson MN (1999) DNA biochip using
a phototransistor integrated circuit. Anal. Chem. 71:358–363.
Wang H, Feng C-D, Sun S-L, Segre CU, Stetter JR (1997) Comparison of conductometric humid-
ity-sensing polymers. Sens. Actuators B 40:211–216.
Wang J (2002) Electrochemical detection for microscale analytical systems: A review. Talanta
56:223–231.
Wang J, Musameh M, Lin Y (2003) Solubilization of carbon nanotubes by Nafion toward the
preparation of amperometric biosensors. J. Am. Chem. Soc. 125:2408–2409.
Wang N, Zhang N, Wang M (2006) Wireless sensors in agriculture and food industry – Recent
development and future perspective. Comput. Electron. Agri. 50:1–14.
Want R. (2004) Enabling ubiquitous sensing with RFID. Computer 37(4):84–86.
Ward M, Buttry DA (1990) In situ interfacial mass detection with piezoelectric transducers.
Science 249:1000–1007.
Wetzl BK, Yarmoluk SM, Craig DB, Wolfbeis OS (2004) Chameleon labels for staining and
quantifying proteins. Angew. Chem. Int. Ed. 43:5400–5402.
Wolfbeis OS (2004) Fiber-optic chemical sensors and biosensors. Anal. Chem. 76:3269–3284.
Wolfbeis OS (2006) Fiber-optic chemical sensors and biosensors. Anal. Chem. 78:3859–3874.
Wu R-J, Sun Y-L, Lin C-C, Chen H-W, Chavali M (2006) Composite of TiO2 nanowires and
Nafion as humidity sensor material. Sens. Actuators B 115:198–204.
Yeo SC, Eisenberg A (1977) Physical properties and supermolecular structure of perfluorinated
ion-containing (Nafion) polymers. J. Appl. Polym. Sci. 21:875–898.
Zahedi S, Navid R, Hassibi A (2004) Statistical modeling of biochemical detection systems. Pro-
ceedings of the 26th Annual International Conference of IEEE Engineering in Medicine and
Biology Society, EMBS; San Francisco, CA, USA, pp. 208–211.
Zemel JN (1990) Microfabricated nonoptical chemical sensors. Rev. Sci. Instrum.
61:1579–1606.
Zen J-M, Kumar AS (2001) A mimicking enzyme analogue for chemical sensors. Acc. Chem.
Res. 34:772–780.
Zhang C, Suslick KS (2005) Colorimetric sensor array for organics in water. J. Am. Chem. Soc.
127:11548–11549.
Zhang Y, Ji H-F, Brown GM, Thundat T (2003) Detection of CrO42− using a hydrogel swelling
microcantilever sensor. Anal. Chem. 75:4773–4777.
Chapter 7
Applications of Functional Thin Films
for Mechanical Sensing
Chang Liu
Abstract This chapter presents the most common fundamental mechanical

transduction principles. Mechanical sensing is contrasted using thin films with
alternative mechanical transduction mechanisms and the pros and cons of using thin
films are touched on. A detailed discussion of displacement and stress development in
various types of structures and loading conditions is included to prepare the reader
for the final section. The final section is dedicated to a case study, where the typical
steps for designing and developing a mechanical sensor are outlined, the materials
selection and fabrication process are illustrated, and the multiple functions that
mechanical sensors can fulfill are highlighted.
Types of Mechanical Sensors and Sensing Principles
Mechanical sensing plays very important roles in today’s world, with applications
such as crash sensors in automotive airbag deployment systems, flow sensors for
monitoring and controlling the flow in semiconductor equipment, blood pressure
sensors and heartbeat monitors in pacemakers, and acceleration sensors for motion-
sensitive electronics gaming controllers. With the growing needs of autonomous
systems and interactive communication and entertainment systems, it is foreseeable
that the needs for high-performance and low-cost mechanical sensors will continue
to grow.
Mechanical sensing is a vast field encompassing many variables, including:
position, displacement, force, stress, strain (deformation), contact pressure, speed,
acceleration, vibration, rotational velocity, fluid pressure, fluid velocity, fluid shear
stress, touch, roughness, texture, and softness or hardness.
C. Liu
Department of Mechanical Engineering, Northwestern University, 2145 Sheridan Road, Rm.,
Evanston, IL60208-3111, USA
e-mail: changliu@northwestern.edu
146 C. Liu
Although there is a large variety of mechanical sensors and a diverse collection

of applications, the principles of sensing can be simplified when considering the
signal transduction pathway. The majority of mechanical sensors transform the
sensing variable into either a displacement or mechanical strain. For example:
• An acceleration sensor employs a proof mass suspended by a mechanical spring
(either a beam or a membrane). The acceleration causes the proof mass to expe-
rience an inertial force, which in turn causes either small displacement or
mechanical strain.
• A pressure or tactile sensor employs a membrane that can be directly contacted
by the object of interest. The contact event causes the membrane to deform. The
displacement or the redistributed strain provides clues about the contact force.
• A flow shear stress sensor may employ a shuttle plate that may experience a
frictional drag force when placed in a flow. The drag force is proportional to the
flow shear stress multiplied by the area of the plate. Flow drag force causes the
shuttle plate to move slightly, thereby providing information about the stress.
Therefore, although there is a wide variety of sensors and applications, the core
design issues for mechanical sensors can be much condensed: to a matter of meas-
uring either displacement or stress accumulation.
There are exceptions, however. A contact sensor may employ a thermally heated
resistive element. Upon contact by an object (say a fingertip), the temperature pro-
file is modified. The change in temperature is reflected into a change of resistance.
This type of thermal-based sensing does not involve mechanically deformable
membranes or beams and does not involve displacement or stress redistribution.
A contact sensor may also employ an inductive coil or capacitive plate. When an
object is near, it will affect the electromagnetic field distribution and therefore
change the capacitance or inductance of the coil. This method does not involve
displacement or stress either.
In this chapter, due to space limitations, we disregard the sensing principles that
do not involve displacement or stress measurement.
Principle of Transduction
Mechanical displacement or stress can be measured in a number of ways. These

methods can be classified into two categories: those that involve functional thin
films and those that do not. Although this book is about functional films and their
use in MEMS, it is important for the readers to gain a more complete picture.
Therefore, a number of representative detection methods that do not involve func-
tional films are briefly introduced.
It is generally difficult to categorically compare the pros and cons of sensors with
and without functional films. Functional films involve more detailed knowledge of
materials preparation. To maintain the properties of functional films, it is often
necessary to maintain dedicated equipment without disruption and disturbances; an
7 Applications of Functional Thin Films for Mechanical Sensing 147
introduction of foreign ions in an ion implantation chamber would damage the

process; the process for sputter deposition of piezoelectric films must be controlled
tightly. It may seem that functional films are more cumbersome. However, there are
certain advantages associated with functional films. To measure the displacement
of a cantilever, for example, one can use a number of means, including electro-
static sensing (nonfunctional film) and piezoresistive sensing (functional film).
Electrostatic sensing has a limited range of motion allowed whereas piezoresistive
sensing can accomplish much greater dynamic range (the ratio of maximum over
minimum responses).
Displacement Measurement Without Functional Thin Films
The displacement of an object can be measured by electrostatic (or capacitive)

means. A capacitor involves two conducting surfaces that are close to each other.
A capacitance is developed between these two surfaces. If one of the surfaces
undergoes motion, the capacitance value will change. The change in capacitance
can be detected electronically using appropriate circuitry.
Capacitive sensing is simple in terms of the materials. It does not involve the use
of functional materials. In most cases, the only requirement is that the two surfaces
be somewhat conductive.
The displacement of an object can be measured thermally. For example, consider
two surfaces that are located close to each other. We designated them as surfaces A
and B, arbitrarily. Suppose one of the surfaces, let’s just say A, is connected to a
temperature sensor. The relative displacement of surfaces A and B will change the
heat transfer pathway between these two surfaces. Suppose one of the surfaces
(either A or B under this circumstance) is heated above the ambient temperature.
The displacement will change the temperature of surface A. The displacement
value can be derived from the temperature variation.
Thermal-transferred based sensing uses simple temperature sensors (such as
resistive temperature sensors) and does not involve functional materials. The advan-
tage is the simplicity of materials systems. Thermal-transfer-based sensing can be
extremely elemental and may not involve moving parts. For example, one type of
vibration sensor marketed by MEMSIC Company uses a heated air pocket to detect
acceleration: under acceleration the heated air pocket will move with respect to the
chip substrate and therefore alter the symmetric temperature profile.
Displacement Sensing with Functional Films
Functional films may occasionally be used for displacement. For example, func-
tional thermal electric-based temperature sensors may be used for measuring tem-
perature changes. Displacement may also alter optical reflectance, transmittance, or
148 C. Liu
the refractive index of the device, and the optical signal can be picked up by
functional thin films.
Stress Sensing with Functional Films
Functional films are widely used for monitoring changes of mechanical strain. This
is the focus of this chapter. Mechanical strain can be detected by using piezoelectric
films or piezoresistive films. As mentioned in Chapter 2, a piezoelectric film exhib-
its the behavior of direct piezoelectric effect, charge accumulation due to applied
stress/strain. A piezoresistive film may be made of an elemental metal, a metal
alloy, doped semiconductor, or organic polymers. Their resistance changes as a
function of applied stress/strain.
Other Uses of Functional Films in Mechanical Sensors
Functional films have other applications for mechanical sensors beyond displace-
ment and stress sensing. These are outlined in this section.
Interestingly, mechanical sensors often involve actuators. Actuators may be used
to actively tune the properties of mechanical elements in order to change the char-
acteristics of the mechanical sensor on the fly, with the purpose of increasing the
sensitivity, increasing the dynamic range, or reducing the noise floor.
Actuators connected with beams or membranes have been used in the past. They
may be based on functional films such as piezoelectric films. A piezoelectric film
also exhibits a reverse piezoelectric effect alongside the direct piezoelectric effect
(for sensing). According to the reverse piezoelectric effect, an applied electric field
changes the stress/strain profile in the piezoelectric film and thereby changes the
position or stiffness of mechanical elements.
Active tuning of sensor characteristics can also be achieved using electrostatic
actuators, thermal bimetallic actuators, or magnetic actuators, to name a few.
Other functional thin films, such as hydrogels, electroactive polymers, and con-
ductive polymers, have also been used for actuation.
Preparation of Functional Thin Films
Functional films can be prepared and integrated into mechanical sensors in a

number of ways, depending on the materials (piezoelectric, piezoresistive, etc.),
the types of substrate (silicon, polymer, etc.), the dimensions of the sensor, and a
number of other factors. A few representative functional thin films commonly
used in mechanical sensors are reviewed here and their methods of preparation
discussed. A detailed discussion of material preparation methods for piezoelectric
and piezoresistive materials can be found in Liu (2005) and Madou (2002).
Piezoelectric Films
Commonly employed piezoelectric materials and their properties are summarized

in Table 7.1. Detailed information about piezoelectric coefficients of selected mate-
rials is summarized in the following.
The most commonly used piezoelectric films nowadays are the PZT material
and the ZnO material. Piezoelectric thin films may be deposited by a number of
means, most notably sputtering and sol–gel deposition.
The lead zirconate titanate (Pb(Zrx,Ti1−x)O3, or PZT) system is widely used in
polycrystalline (ceramic) form with very high piezoelectric coupling. The name
PZT actually represents a family of piezoelectric materials. Depending on the
formula of preparation, PZT materials may have different forms and properties.
Manufacturers of PZT use proprietary formulas for their products.
Techniques that are commonly used for preparing bulk PZT materials (such as
PZT-4 and PZT-5A) are not suited for microfabrication. A number of techniques for
preparing PZT films have been demonstrated, including sputtering, laser ablation,
jet molding, and electrostatic spray deposition (Lu et al. 2003). One of the most
widely used methods to prepare thin-film PZT material for MEMS is sol–gel depo-
sition. Using this method, a relatively large thickness (e.g., 7 mm) can be reached
Table 7.1 Properties of selected piezoelectric materials

Curie
Relative Permitivity Young’s Density Coupling Temperature
Material (Dielectric Constant) Modulus (GPa) (kg/m3) Factor (k) (°C)
ZnO 8.5 210 5600 0.075 **
PZT-4 1,300–1,475 48–135 7500 0.6 365
(PbZrTiO3)
PZT-5A 1,730 48–135 7750 0.66 365
(PbZrTiO3)
Quartz 4.52 107 2650 0.09 **
(SiO2)
Lithium tantalate 41 233 7640 0.51 350
(LiTaO3)
Lithium niobate 44 245 4640 ** **
(LiNbO3)
PVDF 13 3 1880 0.2 80
**Data not available
150 C. Liu
easily (Luginbuhl et al. 1997; Wang et al. 2003), using single- or multiple-layer
deposition.
Using a processing technique called screen printing, even thicker PZT films
can be reached in a single pass (Bernstein et al. 1997; Chen et al. 1994; Walter
et al. 2002), with the highest piezoelectric coupling coefficient being 50 pC/N,
significantly lower than what is achievable in bulk PZT. The screen printing ink
consists of submicron PZT powders obtained commercially, and lithium car-
bonate and bismuth oxide as the bonding agent. After screen printing, the
deposited materials are dried and then fired at high temperature for densifica-
tion. The sol–gel deposition process is constantly being advanced. Pinhole-free
PZT films up to 12 mm thick have been realized with d33 in the 140–240 pC/N
range (Bernstein et al. 1997), although single-layer deposition thickness of 0.1
mm is more common.
ZnO material can be grown using a number of methods, including rf or dc sput-
tering, ion plating, and chemical vapor deposition. In the MEMS field, ZnO is most
commonly deposited by magnetron sputtering (Wenzel and White 1988; Yamamoto
et al. 1980) on various materials, with the c-axis (or Z-axis) close to the normal of
a substrate. For ZnO, the c-axis is spontaneously formed without poling.
Strategies for reducing the intrinsic stress of ZnO have been explored in order to
realize large-area thicker films (Zesch et al. 1991). As-deposited ZnO films have
significant compressive stress, ranging from 1 GPa to 135 MPa (DeVoe and Pisano
2001). The stress can be reduced using thermal annealing (e.g., at 500°C for 5 min)
to the 100 to 80 MPa range.
A popular electrode material on top of the ZnO thin film is aluminum, which can
be etched using a solution of KOH, K3Fe(CN)6, and water (1 g:10 g:100 ml). ZnO
itself can be etched using wet etchants such as CH3COOH:H3PO4:water (1:1:80 ml)
at fast rate (Niu and Kim 2003).
Piezoresistive Films
The most commonly used thin films that exhibit piezoresistive properties are doped
silicon. A schematic diagram of a representative process used for doping selective
regions of silicon with dopant atoms is shown in Fig. 7.1. The desired shape of the
resistor is shown in the topmost Fig.. The resistor feature should be moderately
doped (with concentration ranging from 1015 to 1018 cm−3). The two ends of the
resistor should have higher doping concentration, on the order of 1019–1020 cm−3, in
order to form ohmic contacts with metal leads.
Strain gauges made of thin-film metals do not compare favorably with semi-
conductor strain gauges in terms of piezoresistive gauge factors. However, metal
thin films provide sufficient performance for many applications. Using metal
instead of a semiconductor eliminates the needs of doping and lengthy process
steps. Also metal resistors can be deposited and processed under temperatures
much lower than what would be needed for doping semiconductors. Metal can
contact region
resistor (moderately doped)
with metal leads
(heavily doped)
A A’
desired resistor pattern
dopant atoms mask contact mask

doping
shield doping shield
contact
region
(a) mask heavy doping

A-A’
shield
mask resistor mask
doping mask shield shield doping shield
resistor
(b)
moderate
A-A’ doping
ohmic
metal ohmic
contact mask shield
metal contact
leads
(c)
A-A’
Fig. 7.1 Two-step diffusion doping
generally sustain much greater elongation before fracture. As such, metal resistors
can be placed on polymer materials for polymer MEMS devices (e.g., tactile
sensors; Engel et al. 2003) and provide improved mechanical robustness compared
with silicon counterparts.
Nanocomposite polymer is a new class of promising piezoresistive material. By
doping nonconductive elastomers with conducting particles such as metal powders,
carbon black, and carbon nanotubes, one can turn a nonconducting organic material
into a conducting one through the networks of conducting particles dispersed
uniformly within. A typical resistivity curve as a function of doped conducting
particles is illustrated in (Fig. 7.2). At a percolation threshold, the resistivity of the
matrix drastically decreases.
The goal of designing a successful blend of nanocomposite conductive material
lies in the ability to achieve resistivity transition at a low doping level (or loading
level), thereby avoiding mechanically stiffening the polymer matrix.
152 C. Liu
Tunneling Percolation Ohmic
10
Log resistivity (Ω cm)
2
0.0 0.2 0.4 0.6 0.8 1.0
Fig. 7.2 The resistivity of a nanocomposite elastomer decreases with increasing loading concen-
tration. At the percolation threshold, the resistivity reduces drastically due to formation of the
cooperative conducting network
It has been demonstrated that by doping the polymer matrix with high aspect
ratio carbon nanotubes (multiwall), the percolation limit for resistivity reduction is
on the order of 0.5% by volume. This represents a 20× decrease from the material
loaded with carbon black particles, which have aspect ratios of roughly 1:1.
The increased aspect ratio of CNTS (on the order of 100:1 or higher) significantly
reduces the percolation threshold.
The nanocomposite elastomer is exciting because it is a multifunctional mate-
rial. As shown in Fig. 7.3, the resistance of a resistor formed by the nanocomposite
material may change when a mechanical strain is placed on it. This is because the
mechanical strain changes the concentration and conductive pathways. We have
shown that the effective gauge factor resulting from this piezoresistive behavior can
be as large as 10.
Analysis of Displacement and Stress
Because many physical sensors rely on the measurement of displacement and stress
caused by an external event, it is important to analyze the displacement and stress.
The analysis of beams and membranes under complex loading conditions is suffi-
ciently covered in textbooks such as Gere and Timoshenko (1997). In this chapter
we simply provide the most commonly used formula for analyzing stress and
Fig. 7.3 (Top) Scanning electron micrograph of cross-section of PDMS with exposed CNT tubes.
(Bottom) Schematic diagram of piezoresistivity behavior
displacement of beams and membranes under common, idealistic, loading

conditions.
A beam is a structure member subjected to lateral loads, that is, forces or
moments having their vectors perpendicular to the longitudinal axis. Beams are
usually described by the manner in which they are supported. Boundary conditions
pertain to the deflections and slopes at the supports of a beam. Consider a two-
dimensional beam with movement confined in one plane. Each point along the
length of the beam can have a maximum of two linear degrees of freedom (DOF)
and one rotational degree of freedom. Three possible boundary conditions are sum-
marized below according to their restrictions on DOFs:
1. The fixed boundary condition restricts both linear DOFs and the rotational DOF.
No movement is allowed at the support. At the fixed support, a beam can neither
translate nor rotate. Representative examples include the anchored end of a div-
ing board or the ground end of a flagpole.
154 C. Liu
Table 7.2 Possible Boundary Conditions
2. The guided boundary conditions allow two linear DOFs but restrict the rota-
tional DOF.
3. The free boundary conditions provide for both linear DOFS and rotation. At a
free end, a point on a beam may translate and rotate. A representative example
is the free end of a diving board.
These three distinct types of boundary conditions are graphically represented in
Table 7.2.
Beam Displacement Analysis
The general method for calculating the curvature of a beam under small displace-
ment is to solve a second-order differential equation of a beam:
d2 y
EI = M ( x ), (7.1)
dx 2
where M(x) represents the bending moment at the cross-section at location x and y
the displacement at location x. The x-axis runs along the longitudinal direction of
the cantilever. This second-order differential equation can be solved by the two
boundary conditions at the two ends of a beam. However, due to space limitations,
we do not discuss the solution in detail.
Rather, Table 7.3 summarizes the most commonly encountered loading condi-
tions and the maximum vertical and angular displacement, whereas F is the point
loading force (N, Newton), l the length of beam (m), E the Young’s modulus of
beam material (N/m2), d the vertical displacement (m), and q the angular displace-
ment (arc angle).
Table 7.3 Most Commonly Encountered Beam Loading Conditions and Displacement
End constraints and loading conditions Maximum values of stress and
displacement
Maximum q occurs at the free end
Fl 2
Max value of θ =
2EI
Maximum vertical displacement occurs at the
free end
Fixed-free beam under point loading at the free end. Fl 3
Max. value d=
3EI
q at the free end equals zero due to guided
boundary condition.

free end
Fl 3
Max. value d=
Fixed-guided beam under point loading at the free 12EI
end.
Maximum vertical displacement occurs in the

middle
Wl 3
Max. value d=
192EI
A fixed-fixed beam with point loading applied at
the center.
Maximum q occurs at the end of the beam
Ml
Max q =
EI
free end
Fixed-free beam under a torque loading. The
position of the torque along the length is not Ml 2
Max d=
relevant. 2EI
Beam Stress Analysis
The analysis of intrinsic stress in a beam under flexural loading can be very com-
plicated when the loading condition is complex (e.g., multiple loading, rotational
loading, etc.). Here in this section, we focus on the stress versus loading analysis
for the most common beam, the cantilever, a fixed-free beam under point loading
at the free end.
156 C. Liu
Fig. 7.4 Stress distribution in a uniform and symmetric cantilever beam
When a beam is loaded by force or couples, stresses and strains are created
throughout the interior of the beam. Loads may be applied at a concentrated loca-
tion (concentrated load), or distributed over a length or region (distributed load).
To determine the magnitude of these stresses and strains, one first must find the
internal forces and internal couples that act on cross-sections of the beam.
The distribution of longitudinal stress is first described qualitatively (Fig. 7.4).
Under a transverse loading of a concentrated force at the free end, the torque distribu-
tion through the length of the beam is nonuniform; it is zero at the free end and
reaches a maxim at the fixed end. At any cross-section, the signs of longitudinal
stresses change across the neural axis. The magnitude of stresses at any point on the
cross-section is linearly proportional with respect to the distance to the neutral axis.
The magnitude of the maximum stress associated with individual cross-sections
changes linearly with respect to the distance to the free end, reaching a sectionwide
maximum at the top and bottom surfaces. These are the reasons why piezoresistors
are commonly found on the surface of a cantilever and near the fixed end.
The maximum strain for the entire cantilever occurs at the fixed end, where
x = L. In fact, in many routine design tasks, the sole interest is to find the magnitude
of the maximum stress/strain at the fixed end. The maximum strain is expressed as
a function of the total torque M(x):
M ( x )t FLt (7.2)
e max = = .
2 EI 2 EI
Membrane Displacement and Stress Analysis
Membranes are often used in microsensors. In this section, we review one of the
simplest cases of membrane loading: by a uniformly distributed pressure on one
side.
The governing equation for membrane displacement under a uniform pressure
loading p is:
∂4 w ∂4 w ∂4 w p
+ 2 + = , (7.3)
∂x 4 ∂x 2 ∂y 2 ∂y 4 D
where w is the normal displacement for a point of the membrane at a location

(x, y). The term D represents the rigidity of the membrane. It is related to the
Young’s modulus (E), the Poisson ratio (n), and the thickness of the material (t)
according to
Et 3
D= . (7.4)
12(1 − n 2 )
In the case of a square membrane with fixed boundaries, the two-dimensional distribution
of membrane displacement and the magnitude of longitudinal stress along the x-axis are
illustrated in Fig. 7.5. Several important qualitative observations can be made:
1. The maximum displacement occurs at the center of the diaphragm.
2. The maximum stress occurs at the center points of two opposite edges and in the
center of the membrane. The stress along the edge and the center have different
signs. These locations with high stress value are preferred for placement of
piezoresistive sensors for detecting membrane deformation.
In many application cases, only the maximum displacement and the maximum
stress are of interest. These can be calculated using an empirical formula. The
158 C. Liu
Fig. 7.5 Normalized displacement (left) and stress in the x-axis (right)
σmax
p
a
b wcenter
σcenter
a
a/b 1.0 1.2 1.4 1.6 1.8 2.0

β1 0.3078 0.3078 0.3078 0.3078 0.3078 0.3078 0.3078
β1 0.1386 0.1386 0.1386 0.1386 0.1386 0.1386 0.1386
α 0.0138 0.0138 0.0138 0.0138 0.0138 0.0138 0.0138
Fig. 7.6 Bending of rectangular plate under uniform stress
maximum displacement at the center (wcenter) of a rectangular diaphragm (with

dimension of a × b) under a uniform pressure of p is:
apb 4
wcenter = , (7.5)
Et 3
with the value of the proportional constant a determined by the ratio of a to b.
The value of a can be found by using the look-up table in Fig. 7.6. The maximum
stress (at the center point of the long edge) and the stress in the center of the plate
are:
b1 pb 2
s mas = , (7.6)
t2
b 2 pb 2 (7.7)
s center = ,
t2
with the values of b1 and b2 listed in the table as well.
Examples of Mechanical Sensors
There are many published works that cover mechanical sensors based on functional
thin films. Instead of citing and reviewing each distinct work, we provide one
example of a mechanical sensor that is inspired by the biological haircell. This
example is meant to serve a few purposes:
1. It will outline the typical steps for designing and developing a mechanical
sensor.
2. It will illustrate the fabrication process and material preparation of functional
thin films.
3. It will illustrate the multitude of functions that mechanical sensors can achieve.
My group is developing artificial haircell sensors that mimic the haircell sensor,
widely found in many animals, that performs a large variety of functions. The bio-
logical haircell, a common neuronal mechanoreceptor, is responsible for a wide
variety of sensing in different animal species. A haircell consists of a cilium attached
to a neuronal cell. When the cilium is displaced mechanically, the neuron mem-
brane is placed under stress and the neuron fires electrical pulses, analogous to
digital signals sent through computer bus lines.
Although the structure is exceedingly simple, haircells in biology cover a wide
range of mechanical sensing functions (illustrated in Fig. 7.7). Haircells are respon-
sible for hearing (human cochlea), flow sensing (insects, spiders, and fish), vibra-
tion sensing (insects), equilibrium sensing (human inner ear), and joint angle
sensing (insect), to name a few examples. Since 1998, our group has been developing
artificial haircell sensors as modular building blocks of sensors for flow, vibration,
touch, and acoustic vibration.
Microfabrication and miniaturization technology have been growing rapidly in
the past ten years with the advancement of microelectromechanical systems
(MEMS) technology. MEMS technology offers potentials for realizing high-
performance sensors that contain integrated electronics for on-chip, low noise signal
amplification and efficient signal processing. Sensor performance is uniform as the
photolithography technique ensures high degrees of precision. The technology is
potentially low cost because of the batch processing similar to integrated circuits.
However, the argument for the low-cost aspect of the MEMS technology with
relevance to robotics sensors must be made carefully. The low-cost promise can be
realized if the demand for a particular sensor is high, and hence reduces the costs
of one-time setup and design. Examples include the accelerometers made by Analog
Devices and Motorola. Because the demand for the automotive air-bag is high, the
sensors can be made commercially in such a manner that early research and devel-
opment costs can be compensated.
This scenario is not necessarily true for all sensors that are important for robotics
use. Many sensors, such as flow sensors, tactile sensors, and gyros/balance sensors,
are associated with small commercial demand and unique performance character-
istics. The lack of the economy of scale therefore can prevent applications of advanced
sensors in robotics systems.
160 C. Liu
Haircell
Acoustic sensor
Vibration sensor
Flow sensor
Contact sensor
Fig. 7.7 The many functions of a biological cell
It is important to provide advanced integrated sensors for robotics systems at a

controllable cost. One approach for solving the existing technology bottleneck is to
use modular sensor architecture. Namely, instead of building sensors from the
ground up each time, sensors need to be developed based on a modular design. This
allows the reduction of design and development of the fabrication process, poten-
tially reducing the costs of final sensor products.
Recent development of an artificial haircell sensor is discussed below. The sche-
matic diagram of the sensor is shown in Fig. 7.8a. An artificial cilium, made of high
aspect ratio, photolithographically defined SU-8 hair, is suspended at the free end
of a silicon cantilever. At the base of the cantilever, the silicon is doped selectively by
using boron. When a lateral force acts on the hair, a moment is developed to act on
the horizontal silicon cantilever. The moment introduces longitudinal stress at the
site of the piezoresistor. Naturally, the piezoresistors are located in the region where
the induced stress is the greatest.
Assume a force F is applied in the horizontal direction and at the free end of the
cilium. The height of the cilium is denoted l. The magnitude of the induced strain
(e) at the location of the piezoresistor is:
Flt 6Fl
e= = , (7.8)
2EI Ewt 2
Fig. 7.8 (a) Scanning electron micrograph of an artificial haircell sensor; (b) a schematic
diagram; (c) optical micrograph of an array
where w and t are the width and thickness of the horizontal cantilever beam. The
term I denotes the moment of inertia of the horizontal support beam. The stress
distribution has been simulated using ANSYS finite element software. The result-
ant diagram of stress before and after lateral loading is shown in Fig. 7.9.
The length of the horizontal cantilever beam is on the order of 200–600 mm.
The length of the vertical cilium is approximately the same as the length of the
horizontal beam.
The scanning electron micrograph and optical micrograph of the finished artificial
haircell are shown in Fig. 7.8b,c.
The fabrication process uses a silicon-on-insulator (SOI) wafer that consists of
three layers: a top layer thin epitaxial, single crystal silicon film (approximately 10 mm
thick), a sandwiched silicon dioxide layer (insulator), and a supporting substrate
(500 mm thick) (Fig. 7.10). First, the top epitaxial layer is selected doped into the
shape of a piezoresistor. The doping is completed by using ion implantation of
boron atoms into a silicon lattice. The designed stopping range of the boron atom
is half the thickness of the epitaxial layer.
An oxide insulator is grown over the top surface of the SOI wafer (step b).
The oxide layer is selectively patterned by first depositing a photoresist layer on
top, photolithographically patterning the photoresist, and etching the exposed
silicon oxide through the photoresist mask. The photoresist is then removed by
using organic solvents.
162 C. Liu
Fig. 7.9 Finite element stress analysis
A metal conductor layer is then deposited and patterned. It works wire leads that
connect from the piezoresistors to the periphery of the chip. The wafer is covered
again with a photoresist thin film which is then photolithographically patterned to
form the cantilever. The photoresist serves as a mask to etch the epitaxial silicon to
form the cantilever (d). The wafer’s back side is covered with photoresist and pat-
terned to form deep, through wafer trenches that expose the underlying silicon
oxide layer. The wafer is turned around again. This time the front side is covered
with a thick layer of photodefinable epoxy film called SU-8. The thickness of the
SU-8 layer corresponds to the height of the hair in the end. The SU-8 epoxy is
photolithographically patterned to form the artificial cilium. Finally, the oxide is
removed to allow the cantilever to be free-standing.
Fig. 7.10 Microfabrication process of the artificial haircell
This process is carefully designed such that every step that involves etching
achieves maximum selectivity. In step e, the deep trench etching of silicon has very
low etch rate on oxide, therefore the process is self-terminating. In step g, the oxide
etch has minimum effect on the hair and the substrate.
The artificial haircell sensor can be used to measure vibration, touch, and flow
velocity. For flow velocity measurement, moving fluid past the cilium introduces
frictional force and drag force on the cilium, which in turn generates momentum
and stress distribution at the base of the hair. The haircell sensors have been pack-
aged and tested in water tunnels at various flow speeds. The response of the output
as a function of the input velocity is illustrated in Fig. 7.11. The minimum detection
limit for the sensor is 0.5 mm/s in water.
Conclusions
This section outlines the basic issues, challenges, and opportunities of applying
functional films for mechanical sensors. The types of mechanical sensors were
outlined first. Then the principles that are often used for measuring displacement
and stress within beams and membranes were discussed. Subsequently, the methods
commonly used for preparing functional thin films for integrated sensors were
164 C. Liu
0.8
Sensor Output (V)
0.6
0.4
0.2
0
0.0 0.1 0.2 0.3 0.4
Inflow velocity (m/s)
Fig. 7.11 Output of sensor with respect to water flow velocity
discussed. The basic mechanical analytical formula for estimating displacement

and stress in beams and membranes under prototypical, simple loading conditions
followed. An example was given in the end—an artificial haircell sensor inspired
by biological haircells—to illustrate the typical steps involved for mechanical sensor
development. The typical development process of a physical sensor involves a
number of distinctive stages, such as design, prototyping, packaging, characterization,
calibration, and quality control.
References
Bernstein JJ, Finberg SL, Houston K, Niles LC, Chen HD, Cross LE, Li KK, Udayakumar K
(1997) Micromachined high frequency ferroelectric sonar transducers, IEEE Transactions on
Ultrasonics, Ferroelectrics and Frequency Control, 44:960–969
Chen HD, Udayakumar KR, Cross LE, Bernstein JJ, Niles LC (1994) Development and electrical
Characterization of lead zirconate titanate thick films on silicon substrates, Proceedings of the
Ninth IEEE International Symposium on Applications of Ferroelectrics, University Park,
Pennsylvania, August 7–10, 1994
DeVoe DL, Pisano AP (2001) Surface micromachined piezoelectric accelerometers (PiXLs),
Journal of Microelectromechanical Systems, 10:180–186
Engel J, Chen J, Liu C (2003) Surface micromachined piezoelectric accelerometers (PiXLs),
Journal of Microelectromechanical Systems, 10: 180–186
Gere JM, Timoshenko SP (1997) Mechanics of Materials, 4th edn. PWS, New York
Liu C (2005) Foundations of MEMS. Prentice-Hall, Englewood Cliffs, NJ
Lu J, Chu J, Huang W, Ping Z (2003) Microstructure and electrical properties of Pb(Zr, Ti)O3 thick
film prepared by electrostatic spray deposition, Sensors and Actuators A: Physical, 108:2–6
Luginbuhl P, Collins SD, Racine GA, Gretillat MA, De Rooij NF, Brooks KG, Setter N (1997)
Microfabricated lamb wave device based on PZT sol-gel thin film for mechanical transport of
solid particles and liquids, Journal of Microelectromechanical Systems, 6:337–346
Madou MJ (2002), Fundamentals of Microfabrication: The Science of Miniaturization, 2nd edn.
CRC Press, Boca Raton, FL
Niu MN, Kim ES (2003) Piezoelectric bimorph microphone built on micromachined parylene
diaphragm, Journal of Microelectromechanical Systems, 12:892–898
Walter V, Delobelle P, Moal PL, Joseph E, Collet M (2002) A piezo-mechanical characterization
of PZT thick films screen-printed on alumina substrate, Sensors and Actuators A: Physical,
96:157–166
Wang LP, Wolf RA, Wang Y, Deng KK, Zou L, Davis RJ, Trolier-McKinstry S (2003) Design,
fabrication, and measurement of high-sensitivity piezoelectric microelectromechanical sys-
tems accelerometers, Journal of Microelectromechanical Systems, 12:433–439
Wenzel SW, White RM (1988) A multisensor employing an ultrasonic Lamb-wave oscillator,
IEEE Transactions on Electron Devices, 35:735–743
Yamamoto T, Shiosaki T, Kawabata A (1980) Characterization of ZnO piezoelectric films pre-
pared by RF planar magnetron sputtering, Journal of Applied Physics, 51:3113–3120
Zesch JC, Hadimioglu B, Khuri-Yakub BT, Lim M, Lujan R, Ho J, Akamine S, Steinmetz D,
Quate CF, Rawson EG (1991) Deposition of highly oriented low-stress ZnO films, Proceedings
of IEEE Ultrasonics Symposium, 1:445–448
Chapter 8
Sensing Infrared and Terahertz Regions
by Functional Films
Magnus Willander, Victor Ryzhii, and Qingxiang Zhao
Abstract Designing functional films and nanostructures has a key role in the
performance of the infrared (IR) sensing and terahertz (THz) sensing that are
based particularly on quantum well, wire, and dot structures. Sensing the electro-
magnetic (EM) spectra is an extremely important issue for various fields, from
understanding the universe, living cells, and elementary particles to numerous
applications. To give a glimpse of the field in connection to functional films and
nanostructures as sensing elements, in this chapter we briefly discuss infrared
(IR) sensing and terahertz (THz) sensing. For IR sensing we limit ourselves to
low-dimensional semiconductor functional films. For THz sensing we discuss:
(a) how strain in thin films influences THz absorption from impurities, (b) plasma
effects in two-dimensional electron gas (2DEG), and (c) ultrasensitive bolometers
based on metal films.
Introduction
Sensing the electromagnetic (EM) spectra is an extremely important issue for

understanding the universe to understanding living cells to understanding elemen-
tary particles to numerous applications. To give a glimpse of the field in connection
with functional films and nanostructures as sensing elements, infrared (IR) sensing
and terahertz (THz) sensing are briefly discussed. Infrared radiation covers the
M. Willander () and Q.X. Zhao

Linköping University, Institute of Science and Technology, SE-601 74, Norrköping, Sweden
e-mail: magwi@itn.liu.se
V. Ryzhii
Computer Solid State Physics Laboratory, University of Aizu, Aizu-Wakamatsu, 965-8580, Japan
168 M. Willander et al.
wavelength region of the electromagnetic radiation from 1 to 1000 um. IR radiation

was discovered by Hershel in 1800 and the development of IR detectors started in
the beginning of the nineteenth century. The detectors are usually divided into two
main categories, depending on the interaction between the radiation and the
materials:
• Thermal detectors (thermopile, bolometers, and pyroelectric)
• Photon detectors (intrinsic, extrinsic, photoemissive, quantum well, wire, and
dot detectors)
Designing functional films and nanostructures has a key role particularly for the
performance of quantum well, wire, and dot IR detectors. Below we start with a
thorough analysis of these detectors.
To give a glimpse of the field, infrared sensing and terahertz sensing are briefly
discussed. For IR sensing we limit ourselves to low-dimensional semiconductor
functional films. For THz sensing we discuss three topics: (1) how strain in thin
films influences THz absorption from impurities, (2) plasma effects in two-
dimensional electron gas (2DEG), and (3) ultrasensitive bolometers based on
metal films.
Intersubband Infrared Photodetectors Based

on Quantum Heterostructures
Intersubband Infrared Photodetectors
To detect an infrared radiation signal it can be converted into an electrical signal.

By the principle of operation, infrared detectors can be either thermal or quantum
devices. The operation of quantum, or photon, infrared detectors is associated with
direct variation of electric properties caused by absorption of photons due to quan-
tum transitions between different energy states. The change in electrical conductivity
associated with the absorption of infrared photons can be due to the interband
transitions (between the valence and conduction band), ionization of impurity
states, or the transitions between quantized states (subbands) in the conduction or
valence band. The quantization of electron (hole) states takes place in different low-
dimensional semiconductor structures, in particular, in quantum well (QW) struc-
tures. Intersubband infrared photodetectors are conventionally made of single or
multiple QW structures. Infrared technology based on QW infrared photodetectors
(QWIPs) which utilize the intersubband transitions from QWs has matured rapidly
in the last several years (Ryzhii 2003; Liu et al. 2004; see also references therein).
However, due to dipole selection rules, the intersubband transitions in the conduc-
tion band stimulated by infrared photons polarized in the QW plane are forbidden.
This necessitates the use of different radiation coupling structures in QWIPs, for
instance, gratings.
8 Sensing Infrared and Terahertz Regions by Functional Films 169
The problem of infrared radiation (IR) coupling can be eliminated if electrons in

the intersubband photodetector active region are confined in one or both lateral
directions, in addition to the usual vertical confinement. Such a confinement can be
realized in quantum dot (QD) (Ryzhii 1996) and quantum wire (QR) (Ryzhii et al.
1996) infrared photodetectors (QDIPs and QRIPs, respectively) in which arrays of
QDs and QRs are incorporated instead of QWs. As pointed out previously (Ryzhii
1996; Ryzhii et al. 1996), QDIPs and QRIPs, aside from the sensitivity to normal
incident IR, can exhibit some other features that can be beneficial, for example, an
elevated current gain caused by a reduced capture probability due to the phonon
bottleneck effect and formation of repulsive potential barriers by charged QDs, or
a reduced rate of thermal emission of electrons from QDs and QRs because of an
increased activation energy, among others.
To date, several fabrication and experimental studies of InAs/GaAs, InGaAs/
GaAs, InGaAs/InGaP, and Ge/Si QDIPs have been reported (Berryman et al. 1997;
Phillips et al. 1998, 1999; Kim et al. 1998; Pan et al. 1998, 1999; Maimon et al.
1998; Xu et al. 1998; Horiguchi et al. 1999; Ye et al. 2002; Tang et al. 2002;
Yakimov et al. 1999; Rappaport et al. 2000; Liu et al. 2001; Boucaud et al. 2001;
Miesner et al. 2001). The majority of studied QDIPs are based on QW structures
with vertical electron (hole) transport (perpendicular to the QD arrays). Apart from
QDIP devices, there are successful demonstrations of lateral QDIPs in which elec-
trons propagate parallel to the QD arrays (Miesner et al. 2001; Lee et al. 1999).
Lateral electron confinement is also used in quantum grid infrared photodetectors
(QGIPs) (Rokhinson et al. 1999) and the so-called quantum dot-in-a-well infrared
detectors (QDWIPs) (Raghavan et al. 2002). The principles of QDIP operation,
results of experimental studies of QDIPs, and analysis of their features were
reviewed in some recent publications (Towe and Pan 2000; Ryzhii et al. 2001 a-c,
2004 a-b; Bhattacharya et al. 2002).
In this chapter, we briefly consider and compare QWIPs, QRIPs, and QDIPs
utilizing the analysis based on a semi-quantitative treatment (in line with Ryzhii
et al. (2004 a-b)) of fundamental physical factors determining and limiting the
operation of QWIPs, QDIPs, and QRIPs.
QWIP, QRIP, and QDIP Structures and Principles of Operation
Most QWIPs, QRIPs, and QDIPs are based on vertical heterostructures consisting
of one or several QWs or two-dimensional arrays of QRs or QDs separated by the
barrier layers. The QW, QR, or QD structures serving as the photodetector active
region, where IR radiation is absorbed, are sandwiched between heavily doped
emitter and collector contact layers. The active region can be either doped (with
dopants of the same type as the contact layers) or undoped. Usually the photodetec-
tors in question are made of n+-N-n-N-n+—or N+-N-n-N-N+-heterostructures with
n+—or N+-contact layers, respectively, N-type barrier layers, and n-type QWs, QRs,
(a) Grating (b) (c)

Emitter Emitter Emitter
QWs Collector QRs Collector QDs Collector
QWIP QRIP QDIP
Fig. 8.1 Schematic view of (a) QWIP, (b) QRIP, and (c) QDIP structures (Ryzhii et al. 2004 a-b)
or QDs. Schematic views of vertical QWIP, QRIP, and QDIP device structures are
shown in Fig. 8.1.
The absorption of IR is associated with the electron intersubband transitions
from bound states in QWs, QRs, or QDs into continuum states above the barriers
or into excited quasi-bound states near the barrier top. The bound-to-continuum
transitions or bound-to-quasi-bound transitions followed by fast escape into the
continuum result in the photoionization of QWs, QRs, or QDs and the appearance
of mobile electrons. Bound electrons accumulated in QWs, QRs, or QDs can create
a significant space charge in the active region. In photodetectors made of N+-N-n-
N-N+ heterostructures with the same material of the contact and barrier layers, the
electrons are injected from the emitter to the active region overcoming a potential
barrier in the latter formed solely by the space charge. Hence, the electron injection
in such photodetectors is of thermionic origin.
Due to a conduction band offset at the n+ -N heterointerface in n+-N-n-N-n+
photodetectors, the pertinent heterobarrier prevents the penetration of electrons
from the emitter contact to the active region. However, a sufficiently strong electric
field at this heterointerface, caused by the external electric field and the space
charge, can result in a marked electron tunneling to the active region providing the
electron tunneling injection. The QWIPs fabricated, studied, and used in applica-
tions are primarily based on heterostructures with tunneling injection, for example,
on heterostructures with n+–GaAs contact layers and QWs separated by N-AlGaAs
barrier layers. However, the majority of QDIPs are made of QD structures with the
same material (e.g., GaAs) as the contact and barrier layers, although QDIPs with
more exotic structures were investigated.
Under a bias voltage applied between the emitter and collector contacts the cur-
rent across the active region depends on the applied voltage and the injection condi-
tions (which are dictated by properties of the emitter contact). In normal operation
mode the current is limited by the space charge formed by electrons captured in
QWs, QRs, or QDs. Under dark conditions, the space charge in the active region is
(a) (b)
Emitter
Emitter
QDs (QRs) QDs (QRs)
Collector Collector
Fig. 8.2 Schematic view of the conduction band profiles in single QR array QRIP or single QD
array QDIP with thermionic (a) and tunneling (b) injection. Arrows indicate electron trajectories
in the active regions (Ryzhii et al. 2004 a-b)
determined by the balance between the processes of the electron capture into QWs
(QRs or QDs) and the processes of the electron thermoemission (or tunneling) from
them. The space charge modifies the electric field distribution and affects the bar-
rier at the emitter edge of the active region (in photodetectors with tunneling injec-
tion) or leads to the formation of a controlled potential barrier inside this region.
Under IR illumination, the photoionization of QWs, QRs, or QDs shifts the
balance between the electron capture and emission resulting in the redistribution of
the electric field in the active region which, in turn, gives rise to a change in the
injected current. The total current across the photodetectors includes two compo-
nents: the current caused by the electrons emitted from QWs, QRs, or QDs and the
injected current. Because the capture of mobile electrons is usually a rather slow
process, the main portion of the dark current and photocurrent is due to the injec-
tion. The conduction band profiles in a QRIP and QDIP (with a single array of QRs
and QDs, respectively) and the electron trajectories in their active regions are
shown schematically in Fig. 8.2.
Thus the operation of QWIPs, QRIPs, and QDIPs is associated with the current
across the photodetector active region limited by the bound space charge, which is
controlled by the incident IR radiation. Despite similarities in the QWIP, QRIP, and
QDIP principles of operation, there are some distinctions (Liu et al. 2004; Ryzhii
et al. 2004 a-b):
1. Different degree of the discreteness of the energy spectrum of bound electrons
and, therefore, different statistics of these electrons, capture probability, and
selection rules for intersubband transitions
2. Different spatial distributions of the electric potential in the active region, par-
ticularly in the lateral direction (virtually uniform in QWIPs and strongly non-
uniform in QRIPs and QDIPs with low-density QR and QD arrays)
3. Different dependences of the electron capture probability on the concentration
(number) of bound electrons
The above-mentioned distinctions can result in a marked quantitative variation

of the device performance from one detector type to another, different approaches
to the photodetector optimization, and different applications. The characteristics of
QRIP and QDIP can markedly depend on the lateral dimensions of QRs and QDs.
In QRIPs and QDIPs with QRs and QDs relatively large in the lateral direction
(directions), each QR and QD can have many lateral quantization levels. By con-
trast, if the lateral size of QRs and both lateral sizes of QDs are small, the energy
spectrum of such QRs can comprise only a single one-dimensional energy subband,
whereas the energy spectrum of each QD consists of one discrete quantum level
(shell). Going forward, QRIPs with wide QRs and QDIPs with large QDs are
referred to as L-QRIPs and L-QDIPs, respectively. Similarly, QRIPs with narrow
QRs and QDIPs with small QDs are denoted as S-QRIPs and S-QDIPs.
Comparison of Dark Currents in QWIPs, QRIPs, and QDIPs
The dark current and photocurrent in the photodetectors in question are determined,
first of all, by the rates of thermionic emission from QWs (QR or QD arrays) and
their photoemission by IR. In QWIPs with multiple QWs as well as in multiarray
QDIPs and QRIPs, the contribution of different QWs or arrays can be slightly dif-
ferent. However, for a qualitative study, one can disregard this difference.
The interaction between the electron gas in each QW and the gas of mobile
electrons propagating over continuum states above the inter-QW barriers is rather
weak. Due to this, the energy distributions of electrons in QWs are given by the
Fermi distribution function with the temperature coinciding with the lattice tem-
perature T. At the same time, the electron sheet concentrations in QWs can be far
different from those obtained when the whole electron system (which includes
electrons in QWs, mobile electrons, and electrons in the contacts) is in equilibrium.
In this case, the rate of thermoemission from QWs and QR (or QD) array is deter-
mined by the activation energy ea = ei − eF. Here ei and eF are, respectively, the ioni-
zation energy of the QWs (QRs or QDs) and their Fermi energy (see Fig. 8.3).
The electron gas in QWs can be considered as a two-dimensional system, so
the Fermi energy of electrons in QWs with respect to the bottom of the lowest
subband equals
⎡ p 2 Σ ⎤ p 2 Σ
e F(QW ) = k BTIn ⎢exp( ) − 1⎥ ≈ , (8.1)
⎢ mk BT ⎥ m
⎣ ⎦
where Σ is the average electron sheet concentration in QWs (or, in the follow-
ing, in QR or QD arrays), ¯h and κB are the reduced Planck constant and the
Boltzmann constant, respectively, and m is the electron mass. The last term in
the right-hand side of Equation (8.1) corresponds to the situation where
k BT < p 2 Σ / m .
(a) (b)
εa
εa
εi εi
QR subband QD shells
bottoms
εa εa
εi εi
Fig. 8.3 Energy spectra of (a) QRs and (b) QDs (Ryzhii et al. 2004 a-b)
Each relatively wide QR can have many one-dimensional subbands associated

with the quantization of the energy of electron lateral motion. Such QRs can be
considered as striplike QWs. Disregarding quasi-discreteness of the electron spec-
trum, the Fermi energy can be estimated using the following formula,
e (QW )
⎡
⎢
= k BT In ⎢exp(
p 2 Σ ⎤
⎥
) − 1⎥ ≈
p 2 Σ
mk BT a 2 ∑ QR ⎥ m a ∑ QR
F 2
⎢ (8.2)
⎣ ⎦
−1/ 2
where a and ∑ QR are the width of QRs and the lateral distance between QRs
(lateral period of the QR array), respectively. In the case of QRIPs with rather nar-
row QRs having only one energy level of lateral quantization, the electron system
in each QR constitutes a one-dimensional gas. In this case, one obtains:
Σ
2
p 2
e F(QR ) ≈ . (8.3)
8m ∑ QR
Relatively large QDs (in the lateral directions) can have several quantum shells and
be able to accept a rather large number of electrons. In most experiments, QDIPs
with relatively large QDs were studied. A QD array with QDs of this type can be
considered as a disintegrated QW. Because many quantum shells can be occupied
by electrons in these QDs, one can neglect the discreteness of their electron levels.
In this case, a similar formula to that used for QWs can be used to compute the
activation energy. However, it is necessary to take into account that real sheet
electron density inside the QD equals Σ / a 2 ∑ QD . In the previous expression,
a and ∑ QD are the QD lateral size and the sheet density of QDs in each array,
respectively. The QD density is limited by the condition of a weak overlap of the
wave functions of neighboring QDs (Liu et al. 2004). Otherwise, the probability of
the electron photoemission by normal incident IR can become fairly small; see, for
example, Helm et al. (1994). This means that the product a ∑ QD should not be
2
too close to unity.

As a result, the following expression is obtained.
⎡ p 2 ∑ ⎤ p 2 ∑
e (QD ) ⎢
= k BTIn exp( ⎥
) −1 ≈ . (8.4)
⎢ mk BTa 2 ∑ QD ⎥ ma 2 ∑
F
⎣ ⎦ QD
Both simplified and rather detailed device models of QWIPs with tunneling and
thermionic injection (Liu 1992; Ryzhii 1997; Ryzhii and Liu 1999; Ryzhii et al.
2002 a-b) lead to the following relationship which provides an estimate of the ther-
mal dark current,
eGth
j th ≈ , (8.5)
p
where e is the electron charge, Gth is the rate of thermoemission (per unit area of
a QW, a QR, or QD array), and 〈 p〉 is the average probability of the capture of a
mobile electron passing across a QW.
Considering Equations (8.1), (8.2), (8.4), and (8.5) and introducing ∑ T = mk BT / p ,
2
the density of thermal dark current can be expressed byt
⎡ ⎤
j th(QWIP ) α
1 ⎢exp( ∑ ) − 1⎥ exp(− ei )
p
(QW )
⎢
⎣
∑T ⎥
⎦
k BT
(8.6)
1 ∑ ei
≈ exp( )exp(− )
p
(QW )
∑ T
k BT
a 2 ∑ QR ⎡ ∑ ⎤
e
j (L − QRIP )
α ⎢exp( ) − 1⎥ exp(− i )
⎢ ⎥
∑T a 2 ∑ QR
th (QR )
p k BT
⎢⎣ ⎥⎦
a 2 ∑ QR ∑ e
(8.7)
≈ exp( )exp(− i )
∑ a 2 ∑ QR
(QR )
p k BT
T
a 2 ∑ QD ⎡ ∑ ⎤
j (L − QDIP )
α ⎢exp( ) − 1⎥ exp(− ei )
⎢ ∑ T a ∑ QD ⎥
th (QD ) 2
p k BT
⎣ ⎦
(8.8)
a 2 ∑ QR ∑ ei
≈ exp( )exp(− ).
p
(QR )
∑ T
a 2
∑ QD
k BT
If the densities of QR and QD arrays approach their maxima ∑ QR = a and

−2
∑ QD = a −2 , the distinctions between QW, on the one hand, and QR and QD

(QR ) (QD ) (QW )
arrays, vanish. In this case, one can set p = p = p , and Equations
(8.6)–(8.8), naturally, lead to j th(L − QRIP ) = j th(L − QDIP ) = j th(QWIP ) . Using Equations
(8.3) and (8.5), for rather narrow QRs we arrive at
2
a 2 ∑ QR p ∑ ei
( S − QRIP )
j α exp( exp( − ).
8∑ T ∑ QR
th ( QR )
p k BT (8.9)
In the case of relatively small QDs having a single quantum shell (with maxi-
mum two electrons), considering that e a = e i (see Fig. 8.3), for the thermoemission
rate one obtains:
∑ ei
jth( S −QDIP ) α exp( − ) . (8.10)
∑
( QD )
2 p k BT
QD
Capture Probability
The electron capture probability is determined by many factors: structural and

material parameters of QWs and inter-QW barriers (Rosencher et al. 1994), energy
distribution of mobile electrons (Ryzhii and Ryzhii 1999; Ryzhii et al. 1999;
Kochman et al. 2003), and so on. The energy distribution of mobile electrons affects
the average capture probability because, in part, the dominant capture mechanism
is associated with the emission of optical phonons by electrons. Therefore, a
mobile electron having the kinetic energy exceeding the optical phonon energy
cannot be captured directly. Electron heating results in a decrease of the fraction
of low-energy electrons and, hence, in a decrease in the average capture proba-
bility. Because the energy distribution of mobile electrons is determined by the
electric field, which can give rise to a significant electron heating (Ryzhii and
Ryzhii 1999; Ryzhii et al. 1999; Kochman et al. 2003) with the average kinetic
energy of mobile electrons e >> k BT , the average capture probability exhibits a
steep roll-off with increasing electric field. This, according to Equation (8.10),
results in a steeply increasing dark current–voltage characteristic.
The processes of the electron capture in QRIPs and QDIPs, as already pointed
out, have distinctive features. First of all, the quasi-discreteness and discreteness
of the electron spectra in QRs in QDs can substantially affect the electron
capture. This effect can be particularly pronounced in QDIPs when the quantum-
level separation exceeds the energy of polar optical phonons leading to the
phonon “bottleneck” in the electron capture. Second, due to a limited number of
quantum states in each QD, these states can be fully occupied preventing the
electron capture (because of the Pauli exclusion principle) under certain condi-
tions including excessive doping of the active region and/or large applied voltage.
Third, in QRIPs and QDIPs with low-density QR and QD arrays, the negative
potential of QRs and QDs charged by the captured electrons can result in an
effective repulsion of mobile electrons. This can be a substantial factor limiting
their capture (Ryzhii et al. 2000 a-b, 2001 a-c; Ryzhii 2001; Sergeev et al. 2002).
In particular, in QDIPs, the capture probability can be presented in the following
form (Ryzhii et al. 2001 a-c; Ryzhii 2001),
⎡max N
∑ QD − ∑ ⎤⎦ e ∑
2
∝ (a 2 ∑ QD ) ⎣
( QD )
p exp( − ) (8.11)
max N ∑ QD e CQD ∑ QD
where CQD is the efficient QD capacitance that depends on the QD size as well as
the spacing between QDs. This formula describes the variation of the capture prob-
ability with changing average QD occupancy N = ∑ / max N ∑ QD, where max N
is the maximum number of electrons that can be accepted by a QD. It shows that
p
( QD )
tends to zero when ∑ approaches the maximum value allowed by the
Pauli principle, that is, to max N ∑ QD . Equation (8.11) also shows that p
( QD )
as
a function of ∑ contains an exponential factor associated with the effect of the
repulsion of mobile electrons. Assuming that e corresponds to the temperature T
= 80 K and setting ∑ = ∑ QD , C ≈ 2 ae r ( e r = 12 is the dielectric constant),
QD
p 3/ 2
and a = 15 nm, for the exponential factor in Equation (8.11) one obtains ~ .04. One
can show that the p (QR ) also steeply decreases with increasing ∑ . Equations
(8.6)–(8.9) indicate that the thermal dark current in L-QRIPs and L-QDIPs are fairly
sensitive functions of ∑ and they are steeper than this current in QWIPs. Strongly
∑ and p – ∑ dependences result in an even more
( QD ) ( QR )
decreasing p –
dramatic rise of the thermal dark current in L-QRIPs and L-QDIPs with increasing
∑ and its much higher values compared to those in QWIPs.
T = 80 K
1.5 c = 10
5
1
Ratio of dark currents
1.0
0.5
0.0
0 2 4 6 8 10
Average electron concentration, 1011 cm−2
Fig. 8.4 Ratio of dark current in a S-QDIP and in a QWIP as a function of average electron sheet
concentration at different ratios of capture probability and T = 80 K
c=5 T = 120 K
1.5
80 K
40 K
Ratio of dark currents
1.0
0.5
0.0
0 2 4 6 8 10
Fig. 8.5 The same as in Fig. 8.4 but for different temperatures
In contrast to L-QDIP, S-QDIPs with rather dense QD arrays can exhibit lower
( S − QDIP )
dark current than QWIPs do. Figs. 8.4 and 8.5 show the ratio jth / jth(QWIP ) as
a function of the average electron sheet concentration calculated using Equa-tions
( QW ) ( QD )
(8.6) and (8.10) for different ratios p / p and different temperatures. The
QD density is chosen to be ∑ QD = 5 × 10 cm with the maximum possible number
11 −2
of electrons in QDs equal to max N = 2. Due to closely packed QDs, we neglected

possible weak lateral nonuniformity of the potential of QD arrays. Because of this,
we disregarded the exponential factor in Equation (8.11) and set
= 2c ∑ QD / ⎡2∑ QD − ∑ ⎤ . The quantity c can vary in a fairly
( QW ) ( QD )
p / p
⎣ ⎦
wide range depending on the role of the phonon bottleneck effect in the electron
capture. When this effect can be neglected, one can approximate c ~ 1. As follows
from Figs. 8.4 and 8.5 (see also Ryzhii et al. (2004 a-b)), even at a large para-
meter c, the dark current in S-QDIP can be lower that in QWIPs in a certain range
of the average electron concentrations. However, the dark current in S-QDIPs
becomes rather large when ∑ approaches 2∑ QD because in this case, due to the
( QD )
Pauli principle, p nears zero.
The wetting layers in the QD arrays can markedly affect the electron capture into
QDs. The incorporation of extra barriers between the QD arrays (Ye et al. 2002) can
substantially influence the dynamics and heating of mobile electrons. As a result, the
capture processes in QDIPs with such barriers can be different from those in more
traditional QDIPs. Both the wetting layers and the extra barriers should result in an
increasing capture rate, and, therefore, in a decreasing dark current. Simultaneously
they can result in lowering of the QDIP responsivity (see below).
Responsivity and Photoelectric Gain
Similarly to Equation (8.10), the density of the photocurrent can be related to the
rate of photoemission of a QW (QR or QD array) by the following equation.
eG ph es ∑ I
j ph ≈ = . (8.12)
p p
Here G ph = s ∑ I is the rate of photoemission (per unit area), s is the photo-

emission cross-section, and I is the IR photon flux. According to Equation (8.12),
the responsivity of QWIPs, QRIPS, and QDIPs can, respectively, be presented as
es (QW ) ∑
R ( QWIP )
= ( QW ) (8.13)
Ù p
where hW is the photon energy. Similar formulae can be used for QRIPs and
QDIPs, if s(QW) and (p)(QW) are replaced by s(QR) and (p)(QR), and s(QD) and (p)(QD). One
can see that the rate of photoemission in all the photodetectors under considera-
tion is a rather weak (linear, near linear if the photoemission cross-section depends
somehow on the electron concentration) function of the average electron concentration.
It is instructive that both the dark current and the responsivity do not explicitly
depend on the number of QWs (QR and QD arrays) K in the photodetector.
Under normal operation conditions (p) is rather small. This corresponds to the
situation when the number of electrons emitted from one QW (one array) is much
smaller than the number of the electrons injected by the emitter. In such a case, the
current gain (or photoelectric gain) can be large. This quantity is defined as the ratio
of the total flux of the injected electrons jtk/e to the total rate of thermoemission
from all the QWs Htk = KGtk (under dark conditions) or as the ratio of the total flux
of the injected electrons jpk/e to the total rate of photoemission from all the QWs
Hpk = KGpk (under sufficiently strong illumination), hence, g = jtk/eKGtk or g =
jpk/eKGpk. Using Equation (8.5) or (8.12), the current (photoelectric) gain in QWIPs
can be estimated by
1
g (QWIP ) ≈ ( QW ) (8.14)
K p
For QRIPs and QDIPs one obtains similar estimates. Because the electron capture
processes in QDIPs can be attenuated due to such reasons as the phonon bottle-
neck effect, the Pauli principle, and the formation of repulsive potential, the
responsivity of QDIPs can be substantially higher than the responsivity of
QWIPs. Due to the above-mentioned reasons, the QDIP responsivity can be as
large as several A/W (Raghavan et al. 2002). Relatively large values of the
responsivity are also achieved in QDIPs with lateral structure (Miesner et al.
2001; Lee et al. 1999), in which a small capture probability is primarily due to
repulsion of mobile electrons by charged QD arrays. However, a small capture
probability in QDIPs results in not only high values of the responsivity, but in
large dark current as well.
As shown theoretically in Ryzhii (1997), Ryzhii and Liu (1999), and Ryzhii
et al. (2002 a-b), the electron sheet concentration in QWIPs with different mecha-
nisms of the electron injection from the emitter contact into the active region (tun-
neling or thermionic) is determined by the donor sheet concentration (per one QW)
in this region ∑ D , the number of QWs K, and the applied voltage V. Generally, the
electron sheet concentrations in different QWs in the QWIP can be different. In
QWIPs with the tunneling injection, the electron sheet concentrations in QWs adja-
cent to the emitter can be either smaller or larger than in the QWIP active region
bulk. This is confirmed by numerical modeling of QWIPs (Ershov et al. 1995; Sa’ar
et al. 1998). To estimate the average electron sheet concentration ∑ , one can use
the following simple formula,
2C
∑ − ∑D ≈
e
(V − VC ) (8.15)
where C is a coefficient dependent on the number of QWs K, and VC is some

characteristic voltage. The latter is determined mainly by the electron injection
conditions. This is particularly true in QWIPs with a tunneling injection (Ryzhii
1997; Ryzhii and Liu 1999) VC > 0. However, if the electron injection from the
contact into the QWIP active region is of thermionic origin (Ryzhii et al. 2002 a-b),
VC < 0. Equation (8.15) explicitly indicates that the total charge of the QWIP active
region changes with varying applied voltage. This is because the electric field
induces extra electrons in QWs (Luryi 1985; Ryzhii and Ershov 1995). Similar
calculations (Ryzhii et al. 2000 a-b, 2001 a-c) show that Equation (8.15) can also
be used for QDIPs. Equation (8.15) is in qualitative agreement with experimen-
tal results (Duboz et al. 2003).
Equations (8.6)–(8.9) show that the rate of thermoemission from QWs, QRs, and
QDs exponentially increases with ∑ , which, according to Equation (8.15), is a
function of the applied voltage. As a result, the dark current–voltage characteristics
of QWIPs, QRIPs, and QDIPs are very much in agreement with more detailed
calculations and experimental data. Because the photoemission rate is a much
smoother function of ∑ (what is meant by smoother, lower sensitivity) and,
therefore, V, the responsivity increases with the bias voltage more slowly than the
dark current.
Detectivity
In a thermally limited regime, the detectivity D* of an IR sensor can be expressed

in terms of the total thermoemission and photoexcitation rates Htk = KGtk and Hpk =
KGpk using the following formulae (Ryzhii et al. 2001 a-c; Grave and Yariv 1992).
H ph
D* = (8.16)
2 ÙI H th
The QWIP detectivity can be expressed in terms of the average electron concentra-
tion ∑ by
s QW K ∑ ei
D*(QWIP ) α exp( )
2k BT
∑
exp( ) −1
∑ T (8.17)
∑ ei
≈ s QW K ∑ exp( −
2∑ T
) exp(
2k BT
).
The detectivities of L-QRIPs and S-QRIPs are given by the following equations,
respectively.
ei
s QR K ∑ exp(
2k BT
)
D*( L −QRIP ) α
(a 2 ∑ QR )1/ 4
∑
exp( ) −1
∑ T
a 2 ∑ QR
(8.18)
s QR K ∑ ∑ ei
≈ exp( ) exp( )
(a 2
∑ QR
1/ 4
) 2∑ T a 2
∑ QR
2k BT
2
s QR K ∑ p2 ∑ ei
D*( S −QRIP ) α exp( ) exp( ).
(a 2
∑ QR
1/ 4
) 16∑ T ∑ QR 2k BT (8.19)
For the detectivities of L-QDIPs and S-QDIPs one obtains, respectively:

ei
s QD K ∑ exp(
2k BT
)
D*( L −QDIP ) α
(a 2 ∑ QD )1/ 4
∑
exp( ) −1
∑ T
a 2 ∑ QD (8.20)
s QD K ∑ ∑ ei
≈ exp( ) exp( )
(a 2
∑ QD ) 1/ 4
2∑ T a 2
∑ QD
2k BT
2
s QD K ∑ p2 ∑ ei
D*( S −QDIP ) α exp( ) exp( ). (8.21)
(a 2
∑ QD
1/ 4
) 16∑ T ∑ QD 2k BT
Relationships (8.17)–(8.21) show that the detectivity of detectors under consid-

eration is proportional to K ; that is, the detectors with a large number of
QWs (QR or QD arrays) exhibit higher detectivity (Choi 1997). Using Equations
(8.17)–(8.20), one can obtain expressions for the maximum values of the detec-
tivity at a given temperature attained at certain values of ∑ , that is, at certain
doping levels and applied voltages (generally different for different types of
photodetectors):
ei
max D*(QWIP ) α 2 K s QW ∑ T exp( − 1) (8.22)
2k BT
ei
max D*( L −QRIP ) α 2 K s QR (a 2 ∑ QR )1/ 4 ∑ T exp( − 1) (8.23)
2k BT
8∑ T ∑ QR ei 1
max D*( S −QRIP ) ∝ s QR K exp( − )
p (a 2 ∑ QR )1/ 4 2k BT 2 (8.24)
ei
max D*( L −QDIP ) ∝ 2s QD K ∑ T a 2 ∑ QD exp( − 1).
2k BT
(8.25)
As follows from expression (8.21), the detectivity of S-QDIPs monotonically

increases with increasing ∑ . However, the latter quantity is limited due to the
Pauli principle by the value 2∑ QD . As mentioned above (see also Liu
et al. (2004)), the current gain can dramatically increase when ∑ approaches
2∑ QD leading to a very large dark current and responsivity simultaneously. Never-
theless, to estimate D*(S−QDIP) in such QDIPs, we set ∑ = 2∑ QD in Equation (8.21).
As a result, we obtain:
ei
max D*( S −QDIP ) ∝ 2s QD K ∑ T ∑ QD exp( ).
2k BT (8.26)
Relationships (8.22), (8.23), and (8.25) yield the following expressions for the
ratios of the detectivities.
max D*( L −QRIP ) s QR 2

≈ (a ∑ QR )1/ 4
max D*(QWIP ) s QW (8.27)
max D*( L −QDIP ) s QD

≈ a 2 QD . (8.28)
max D*(QWIP ) s QW
Simultaneously, relationships (8.24) and (8.26) lead to:
max D*( S −QRIP )

≈ 0.8
s QR 2
(a ∑ QD )1/ 4
∑ QR
(8.29)
max D *( QWIP )
s QW ∑ T
and
max D*( S −QDIP )

≈ 2.7
s QD ∑ QD
QD
max D *( QWIP )
s QW ∑ T (8.30)
The dependences of the QWIP, L-QDIP, and S-QDIP detectivities (normalized by

factor exp(ei/2kBT)) on the average electron sheet concentration per one QW and
one QD array calculated using formulas (8.17), (8.20), and (8.21) are plotted in
T = 80 K (a)
4 S-QDIP
Normalized detectivity, a.u.
0
T = 40 K QWIP (b)
6
L-QDIP
4
0
0 2 4 6 8
Fig. 8.6 Normalized detectivities of QWIP, L-QDIP, and S-QDIP versus average electron sheet
concentration at (a) T = 80 K and (b) T = 40 K
Fig. 8.6. We set for simplicity sQW = sQD. It is assumed that ∑ QD = 1 × 10 cm in

11 −2
an L-QDIP and ∑ QD = 5 × 10 cm (so the product a ∑ QD is approximately the

11 −2 2
same for both QDIPs). Fig. 8.6 shows that even if the QD density is markedly
smaller than the maximum possible density (see below), the detectivity of S-QDIPs
can significantly exceed that of QWIPs (Ryzhii et al. 2004 a-b). The superiority of
S-QDIPs over QWIPs in detectivity can be particularly pronounced at low
temperatures.
Let us estimate max ∑ QD / ∑ T which determines the ratio of the S-QDIP and
QWIP responsivities. The QD density in QDIPs is limited by the requirement of a
weak overlap of the wave functions of neighboring QDs (Liu et al. 2004). Due to
such an overlap, a narrow miniband can arise. Its width in QDIPs is estimated as
1 2 me i
D ≈ e i exp( − ).
2 ∑ QD
(8.31)
Broadening of the ground states in QDs into the miniband does not affect the
activation energy if D << 2k BT . Taking into account this inequality and using
Equation (8.31), we obtain the following condition.
2 me i
∑ QD
<
e
. (8.32)
2 ( In( i ))2
2k BT
Condition (8.32) can be rewritten as
∑ QD e
< 2p ( i ) 2
1
. (8.33)
∑ T
k BT In (e i / 2k BT )
Using Equation (8.30) and condition (8.33), one can obtain the following
inequality,
max D*( S −QDIP ) s QD e i 1 s QD

*( QWIP )
< 17( )( ) 2 =( )G (8.34)
max D s QW k BT In (e i / 2k BT ) s QW
s QD
in which the right-hand side can be fairly large at not too small ratio because
s QW
G >> 1 . Indeed, choosing ei = 100 meV and T = 40 – 80 K, one can obtain T ~
390–425. Taking into account relationship (8.34), one may conclude that S-QDIPs
with QD densities close to the maximum value can really exhibit much larger detec-
tivity than QWIPs if the photoemission cross-section sQD for normal IR incidence
and sQW for inclined incidence are close to each other or, at least, the former is
not very small. However, one needs to note that the ratio of the photoionization
cross-sections can strongly depend on the structure of QR and QD arrays (Vasanelli
et al. 2001; Li et al. 2003).
Conclusions
We considered the operation principles of QWIPs, QRIPs, and QDIPs and features
of their characteristics. The comparison of these infrared photodetectors showed:
• QRIPs and QDIPs can exhibit much higher responsivity than QWIPs due to
lower capture probability and, therefore, larger photoelectric gain. Higher
responsivity is inevitably accompanied by higher dark current, because it is
amplified with the same gain.
• QRIPs and QDIPs based on low-density arrays of relatively large QRs and QDs
(L-QRIPs and L-QDIPs, in our terms) should definitely be inferior to QWIPs in
detectivity.
• QRIPs and, particularly, QDIPs based on extremely dense arrays of narrow
QRs and small QDs, in which the bound electrons are really one-dimensional
and zero-dimensional, respectively, can significantly surpass QWIPs in
detectivity.
Sensing THz Region
The terahertz frequency regime is defined between approximately 1 mm (300

GHz) and 100 um (3 THz). THz technology has its background from molecular
astronomers and chemical spectroscopists. Today many new application areas
are coming up to use the THz technology. The term terahertz first became
popular in the mid-1970s. THz sensors have the fastest development of all
submillimeter-wave components and they are nearly quantum limited detectors
up to several THz.
The detectors operate as
• Heterodyne semiconductor components
• Heterodyne superconductor components
• Direct detectors
Heterodyning is used for passive components to increase signal-to-noise ratio by
reducing the bandwidth. For semiconductor components usually a Schottky barrier
is used. Superconducting components in heterodyne detection are usually based on
the Josephson effect, a superconducting–semiconducting barrier, or bolometric
devices. A lot of attention is given today to direct detection techniques and compo-
nents such as bismuth-coated suspended micromachined silicon, superconductor–
insulator–normal metal junction bolometers or the quantum dot single-photon
detector.
Below we discuss two different detectors:
• The semiconductor plasma wave detectors (utilizing that plasma waves propa-
gate with much higher velocities than electrons and can operate up to room
temperature)
• The cold-electron bolometer, ultrasensitive and operating at low temperature
Both devices are examples of what can be obtained in THz sensing by using func-
tional films and nanostructures. But first we discuss impurities in QW structures for
THz transitions.
Impurity in Quantum Well Structures
The shallow donors and acceptors in semiconductors are the most basic impuri-
ties, which control the electrical properties of materials. Understanding the elec-
tronic structures of such impurities is very important in semiconductor physics
and in applications of semiconductor devices. The impurities in bulk semicon-
ductor materials have been investigated for a long time and their properties are
well established. Due to the recent development of advanced growth techniques
such as molecular beam epitaxy and metal–organic chemical vapor deposition, it
is possible to fabricate ultrasharp interface semiconductor heterostructures, such
as quantum well and superlattice structures. Such low-dimensional semiconductor
heterostructures led to a revolution in semiconductor device applications.

Therefore, the investigation of impurities confined in low-dimensional semicon-
ductor structures is important.
In this review, we discuss the progress in the study of acceptors confined in
quantum wells. The emphasis is on the experimental results in comparison with
proper theoretical calculations for the electronic structures of acceptors confined in
strain-free and strained quantum well (QW) systems.
Confined Acceptors
When a semiconductor layer with small bandgap energy is sandwiched between

two semiconductor layers with large bandgap energy, the charge carriers (electrons
or holes) cannot move freely through the different layers anymore. The restriction
of the motion of the charge carriers in such a direction (hereafter referred to as the
z-direction) may be viewed as carrier confinement in a one-dimensional potential
well. Consequently, the motion of the particles in the z-direction is quantized,
whereas their motion in the x- and y-directions does not have any restriction, and
may be treated the same as it is treated in a three-dimensional crystal. Such a
system with restriction of motion in one direction is often referred to as a two-
dimensional quantum well structure.
Because in quantum well structures the translational invariance is not valid along
the growth axis (z-direction), the impurity binding energy depends explicitly on the
precise location of the impurity. In addition, the binding energy of impurity depends
on the characteristic dimension of the well. In Fig. 8.7 we illustrate a few cases,
where the circles represent the cross-section of the impurity Bohr radius of the 1S
impurity wavefunction. When the impurity is located at the center of the well layers,
if the well thickness L >> a0, the quantum well confinement will not introduce any
significant influence on the acceptor wavefunction resulting in no change in the
impurity binding energy. In contrast, when L is comparable or smaller than a0, the
L>>a0 L~a0
Fig. 8.7 Schematic drawing of a quantum well potential. The circles represent the Bohr radius of
the acceptor wavefunction for two different cases
confinement potential will compress the acceptor wavefunction along the growth
direction, resulting in the modification of the impurity energy levels and hence the
modification of the Bohr radius of the 1S wavefunction. The detailed theoretical
treatment concerning acceptors is presented later. Here we briefly discuss some
characteristics of acceptors confined in QW structures.
Near the G point in the Brillouin zone, the upper valence band consists of three
p-like states and is consequently sixfold degenerate, if the spin is taken into
account. Due to the spin–orbit interaction, these states are separated into two
groups, which are characterized by the total angular momentum quantum number
J = 1/2 and J = 3/2, with the J = 3/2 band being lowest in energy. The corresponding
acceptor ground state is denoted 1S3/2 (G). This acceptor level is fourfold degenerate
in the bulk case, but splits into two Kramers doublets, with the heavy hole and light
hole characters in the quantum well case due to lower symmetry. The electronic
structure of the acceptor bound exciton (BE) is illustrated in Fig. 8.8. Depending
on whether the electron–hole interaction or the cubic crystal field is the strongest
effect, the acceptor BE states can be denoted according to the j–j coupling theory;
that is, the BE states are j = 1/2, 3/2, and 5/2, or the symmetry from the crystal-field
scheme, that is, G6, G8, and G7 + G8. In the following discussion, we denote the
acceptor BE as j = 1/2 (G6), j = 3/2 (G8), and j = 5/2 (G7,8), respectively. The transi-
tion A1 (B1) of the acceptor BE corresponds to the transitions between the j = 5/2
(G7,8) (j = 3/2 (G8)) BE state and the heavy hole (hh) acceptor state (1S3/2(G6)),
whereas A2 is related to the j = 5/2 (G7,8) BE and the light hole (lh) acceptor state
(1S3/2(G7)).
e-h coupling
h-h (Crystal-field
Acceptor BE QW potential
coupling scheme)
−
J=0,S=1/2 J=1/2 (Γ6)
+
+ − J=2,S=1/2
J=3/2 (Γ8)
J=5/2 (Γ7,8)
A1 A2 B1 B2
Acceptor 1S3/2 (Γ6)

+ J = 3/2
−
1S3/2 (Γ7)
Fig. 8.8 Schematic picture of the acceptor bound exciton (BE) and the acceptor states at different
perturbation conditions. The possible acceptor BE transitions confined in QW structures are also
indicated in the figure
Theory of Confined Acceptors in QW
As mentioned above, the calculations of the acceptor electronic structures are more
complicated than the donor case, due to the multiple-valence band nature and
strong Coulomb coupling between the different subbands. The QW potential lifts
the degeneracy of the valence band at G8 into two doublets with symmetry G6 and
G7, due to lower symmetry from Td in bulk material until D2v in symmetric QW
structures. When acceptors are introduced at the center regions of the well layer, the
symmetry is unchanged, so the ground states of the acceptors also split into two
doublets. However, if doped acceptors are located away from the center region, the
symmetry from the D2v in the center doping case reduces to C2v symmetry.
The theoretical calculation of acceptors confined in QW structures are mainly
based on either the effective mass theory (Masselink et al. 1983, 1985; Pasquarello
et al. 1989; Loehr et al. 1990; Fraizzoli and Pasquarello 1990, 1991) or the tight-
bindinglike model (Einevoll and Chang 1990) in strain-free QW systems such as
GaAs/AlGaAs structures. The calculations (Masselink et al. 1983, 1985; Pasquarello
et al. 1989; Einevoll and Chang 1990) provide the energies of the acceptor ground
and the excited states. It is found that the binding energy of the acceptor ground
state G6 depends on the well width and the location of acceptors in the well layer.
At a later stage, the effective mass calculations of acceptor energies, based on the
work by Pasquarello et al. (1989), were expanded to include magnetic field effects
and stress effects or built-in strain in a lattice-mismatched QW system by QX (Zhao
et al. 1994a, b1998; Zhao and Willander, , 1999, 2000; Zhao et al. 2001). Based on
these calculations, the detailed energy levels of acceptors and oscillator strengths of
the transitions between the acceptor ground and different excited states in QW
structures in the presence of magnetic field or stress perturbations were obtained
and can be compared with experimental data.
The acceptor Hamiltonian expressed in electron energy is given by the following
operator,
(
H = − H kin + H c + H QW ) (8.35)
⎡P + Q L M 0 ⎤
⎢ L+ P −Q 0 M ⎥⎥
H kin =⎢ + (8.36)
⎢ M 0 P −Q −L ⎥
⎢ + + ⎥
⎣ 0 M −L P + Q⎦
where
g 12 2
P= k
2 m0
g 2 2 2
Q= (k x + k y2 − 2 kz2 )
2 m0
g 3 2
L = −i 3 (k x − ik y )kz
2 m0
2
M= 3 (g 2 + g 3 )(k x − ik y )2 .
4 m0
The Hc and HQW are the Coulomb potentials due to the acceptors and the quantum
well potential, respectively. The QW potential, HQW, must properly include the
effects of the deformation potential and bandgap offset between well and barrier
materials. That is, HQW will contain a square-well potential, DEQWhh,lh, for the hh and
the lh, and includes a potential difference (Vp) between the hh and lh band edges in
the well due to built-in strain, thus:
H QW = DEhh
QW
,lh + Vp .
Once all effects are properly included in the acceptor Hamiltonian, the Hamiltonian
given in Equation (8.35) acts on a four-component function:
F m ( r, q , z ) = ⎡⎣ F m , s ⎤⎦ = ⎡⎣ F m ,3/ 2 , F m ,1/ 2 , F m , −1/ 2 , F m , −3/ 2 ⎤⎦ .
The energy levels of the shallow acceptor states and corresponding wave functions
are derived. The s component of an acceptor envelope function of definite angular
momentum m can be expanded into a set of basis functions, separable in the coor-
dinates r and z:
F m , s ( r, q , z ) = ei ( m − s )q ∑ Anlm , s r m − s e −al r gns ( z ).
By using these wavefunctions, most of the integrals that appear in the matrix ele-
ments of the Hamiltonian can be computed analytically. Particularly, in the calcula-
tion of the matrix elements of the Coulomb potential Hc, an auxiliary integral that
decouples r and z coordinates is introduced using the well-known transformation:
1
= ∫ e − z q J 0 ( rq )dq.
(r − z )
2 2
In the following, we present an example of the experimental results in comparison

with the corresponding calculated results in order to demonstrate the validity of the
previously described theoretical model.
When an external field (electrical, magnetic, or uniaxial strain) is applied, the
above calculations have to be modified properly in order to take into account external
field effects. For more details we refer to the review book by Holtz and Zhao
(2005).
Information concerning the electronic structure of acceptors can be experi-
mentally obtained from optical spectroscopy such as infrared absorption,
photoluminescence, and Raman scattering measurements. The infrared absorption

measurements, for example, by Fourier transform infrared (FTIR) spectroscopy, is
a direct way to measure the energy separations between the acceptor ground and
the different p-like excited states. The rich information concerning the electronic
structures of acceptors from infrared absorption measurements has been demon-
strated in bulk GaAs, where the far-infrared photoconductivity measurements were
used. From these measurements, the transitions 1S3/2–2P3/2, 1S3/2–2P5/2 (G8), and
1S3/2–2P5/2 (G7) were observed and their corresponding transition energies have been
experimentally deduced. Consequently, detailed information on the electronic
structure of acceptors can be achieved this way. However, the situation for quantum
well structures is more difficult. Due to a small absorption volume of the acceptor
doped quantum well structure, the infrared measurements are difficult to perform
and the derived experimental information is still limited.
A more complete review of the experimental work can be found in a recently
published book by Holtz and Zhao (2005). In this chapter, we focus on the influ-
ence of built-in strain on the acceptor states in InGaAs/AlGaAs QW structures.
When increasing the In-concentration, the separation between the lh and the
hh states increases due to the built-in strain effects. The solid lines in Fig. 8.9 are
calculated results and dotted plots represent experimental data (Zhao et al. 2000).
Both the tendency and absolute energy values of the acceptor states with
In-concentration show an excellent agreement between the experimental and the
25.00
2S-1S
20.00 2P-1S
Energy (meV)
15.00
Δ (lh-hh)
10.00 Exciton
5.00
2P-2S
Δ(1S)
0.00
−0.002 0.000 0.002 0.004 0.006 0.008 0.010
In concentration
Fig. 8.9 The acceptor 1S-2S and 1S-2P transitions versus In-concentration in Be-acceptor doped
InxGa1-xAs/Al0.3Ga0.7As QW structures. Dots are experimental data and lines are the theoretical
calculated results according to the model discussed in text. ΔExciton(lh–hh) represents the energy
separation between the light and heavy hole-free exciton transition. Δ(1S) is the energy separation
between the acceptor 1S3/2(G6) and 1S3/2(G7) states
calculated results. We would like to point out that according to the theoretical
calculations, the change of the acceptor electronic structures, in the range of
In-concentrations used here, is mainly due to the variation of the biaxial potential.
Based on the previous examples, we can confidently conclude that the used theo-
retical model is very reliable in calculating the energy levels of acceptors confined
in QW structures. Concerning the possible stimulated far-infrared emissions, the
research has touched on many different ways (Brazis 1995; Andronov 1987; Faist
et al. 1994; Sirmain et al. 1997). Particularly, a far-infrared stimulated emission
from uniaxially stressed p-Ge (Odnoblyudov et al. 1998, 1999; Altukhov et al.
1992) reported recently has become very interesting. On the other hand, the pos-
sible THz emission can also be used for the reversed process, that is, THz detection.
According to estimated critical layer thickness (Matthews and Blakeslee 1974;
Tsao et al. 1987; People and Bean 1985; Ekenstedt et al. 1993; Wang et al. 1991;
Fritz 1987), the possible THz transitions have been calculated in the strained QW
systems such as InGaAs/GaAs and SiGe/Si QW structures (Zhao and Willander
2000). Therefore, the THz sensing devices based on impurity states in QW struc-
tures can be designed according to the above theoretical calculations.
Plasma Waves and Oscillations in Two-Dimensional

Electron Systems
There is a strong demand in compact semiconductor terahertz devices. Such devices

include, in particular, detectors, frequency multipliers, and coherent and tunable
sources of THz radiation. Conventional classical semiconductor devices such as
bipolar transistors, field-effect transistors, Gunn diodes, diodes utilizing electron-
transit-time effect, and so on, cannot reach the THz range or are rather ineffective
in this range of frequencies. A significant portion of the THz range, known as the
THz gap, is still not covered by compact and effective devices.
The utilization of specific traveling and standing plasma waves in two-dimensional
electron gas (2DEG) systems based on semiconductor heterostructures appears to be
rather promising (Dykonov and Shur 1996; Ryzhii 2003) for THz devices. As an exam-
ple of a heterostructure with 2DEG, one can refer to a field-effect high-electron-
mobility transistor (HEMT) structure schematically shown in Fig. 8.10a. The point is
that in contrast with 3DEG semiconductor structures, 2DEG systems can exhibit rather
high electron mobility and, hence, relatively low electron collision frequencies. This can
provide weak damping of plasma oscillations in 2DEG systems that promote pro-
nounced plasma-wave effects. Apart from this, the spectra of plasma waves in 2DEG
systems are markedly different from those obtained in 3DEG systems (Stern 1967;
Chaplik 1972; Nakayama 1974) and the frequencies of 2DEG systems with easily
achievable parameters can fall into the THz range (Dykonov et al. 1996; Shur and
Ryzhii 2003).
This chapter deals with the review of concepts associated with applications of
plasma waves and oscillations for different compact and effective THz devices.
(a) Gate layer (structure)

Source Gate Drain
W
+ + + + + + + + + +
– – – – – – – – – –
2DEG
channel
Lc Lg Lc
High electric field

(b) gate-drain region
Source Gate Drain
ΔLg
Ld
(c) (d) Gate RT structure

Gate layer
2DEG
channel Gate
Gate
Fig. 8.10 Schematic view of HEMT structure (a) at moderate drain voltage and (b) in saturation
regime (high drain voltage); (c) band diagram of HEMT doped gate layer; and (d) band diagram
of HEMT with a resonant-tunneling structure between the channel and the gate instead of the gate
layer. Arrows show electron tunneling from the channel into the gate
A fairly long history of attempts to use plasma-wave effects in 3DEG systems

and their extensive theoretical and experimental studies (see, e.g., Kustov et al.
(1980), Bannov and Ryzhii (1983), Ryzhii and Fedirko (1983), Ryzhii et al.
1984, Hu and Wilkins (1991), Kempa et al. (1993), and Kersting et al. (1997))
have not yet materialized into the creation of THz devices with the required
characteristics (although some old ideas can be invoked to explain experimental

findings related to 2DEG systems). Even though the special features of plasma
waves in 2DEG systems have been known for decades (Stern 1967; Chaplik
1972; Nakayama 1974) and these waves were observed experimentally a long
time ago (Allen et al. 1977), the practical interest in device applications was, in
our view, first stimulated by Dyakonov and Shur’s paper published in 1993
(Dyakonov and Shur 1993). In this paper, the authors proposed a novel mecha-
nism of self-excitation of plasma waves (plasma instability) in HEMTs. In light
of this development, our attention is directed to plasma-wave effects just in
transistorlike 2DEG systems discussing primarily recent device proposals.
Because the electron concentration in 2DEG systems in most interesting hetero-
structures and devices based on these heterostructures is rather large, the electron–
electron interactions play an important role. To increase the electron mobility in
2DEG, the donors are usually placed at some distance from the 2DEG channel so
the channel and the doped layer are separated by a spacer layer. As a result, the
frequency of electron–electron collision can markedly exceed the frequency of
electron collisions with the donors, residual impurities, and phonons. In this case,
2DEG systems can be described to a good approximation by a hydrodynamic elec-
tron transport model.
The equations of this model comprise the Euler equation and the continuity
equation:
∂u ∂u e ∂y
+ + nu = ⏐z = 0 (8.37)
∂t ∂x m ∂x
∂∑ ∂∑ J
+ = . (8.38)
∂t ∂x e
Here, ∑ ( x, t ) are the hydrodynamic electron velocity and the electron sheet
concentration, respectively, y(z,x,t) is the self-consistent electric potential in the
area surrounding the 2DEG channel, u is the collision frequency of electrons with
impurities and phonons (not the electron–electron collision frequency), e and m are
the absolute value of electron charge and the electron effective mass, J = J(x,t) is
the leakage electron current from the 2DEG channel or the electron current injected
into the channel, the x-axis is directed in the 2DEG channel plane, and the z-axis is
directed perpendicular to this plane. Equation (8.37) can be generalized by intro-
ducing the terms associated with the electron pressure and the 2DEG viscosity.
However, these terms are not particularly essential (Dyakonov and Shur 1996;
Rudin and Samsonidze 1998; Rudin et al. 1999). Equations (8.37) and (8.38)
should be supplemented by the Poisson equation for the self-consistent electric
potential (Dyakonov and Shur 1996; Shur and Ryzhii 2003; Nakayama 1974):
∂ 2y ∂ 2y 4pe
+ = (∑ d − ∑ ) · d ( z ), (8.39)
∂z 2 ∂x 2 k
where k is the dielectric constant, ∑ is the donor sheet concentration, and d(Z)
d
is the Dirac delta function. In the case of a 2DEG channel with a highly con-
ducting plane (gate) placed parallel to the channel at a sufficiently short dis-
tance W (in HEMT-like structures, W is the thickness of the gate structure
between the channel and the gate contact), the above 2D Poisson equation can
be replaced by its simplified consequence (the so-called gradual channel
approximation; Shur (1990)):
j 4pe
= (∑ − ∑ d ). (8.40)
W k
Here, f=f(x,t) is the electric potential in the 2DEG channel; that is, j = j z = 0 .
Equations (8.37)–(8.40) are widely used for theoretical analysis of plasma effects
in 2DEG systems.
To study plasma wave propagation and the response of the 2DEG system to
external perturbations (incoming signals) a small-signal analysis based on the
linearized versions of Equations (8.37)–(8.40) is used. In this approach, the ac
components of the electron hydrodynamic velocity uw ( x ) exp( −iwt ), the electron
sheet concentration, and the self-consistent potential are presented as
uw = uw ( x ) exp( −iwt ), ∑ w = ∑ w ( x ) exp( −iwt ), and jw = jw ( x ) exp( −iwt ), respec-
tively, where w is the signal frequency. In particular, these linearized equations
obtained for a uniform 2DEG channel with the dc electron sheet concentration
∑ 0 result in the following dispersion equations for the plasma waves with
the frequency w and the wavenumber q propagating in the x-direction (so that
jw α exp(i(qx − wt )) ),
2pe2 ∑ 0
w (w + in ) = q (8.41)
km
in the ungated 2DEG channels, and
4pe2 ∑ 0 W
w (w + in ) = q2 (8.42)
km
in the gated 2DEG channels with W < < Lg (Lg is the length of the gated section, i.e.,
the gate length). Equations (8.41) and (8.42) applied to 2DEG channels with
relatively small collision frequency ( n << w ) result in Rew α ∑ 0 q (Stern 1967)
and Rew α ∑ 0 Wq (Chaplik 1972; Nakayama 1974), respectively. The spec-
tra of plasma waves in both ungated and gated 2DEG channels are strongly different
from the spectrum of plasma waves in 3DEG, where Re w is virtually independent
of q. The damping of plasma waves in channels of both types is determined mainly
by the electron collision frequency: w ≈ n / 2 . The electron leakage from (injection
into) the 2DEG channel due to tunneling or resonant tunneling (RT) through the
gate layer or more complex structure (see Fig. 8.12) can lead to an additional damp-
ing or, vice versa, to its suppression. In some cases, the latter processes can result
in a negative damping, that is, in plasma wave instability.
The 2DEG channels usually have limited length and are supplied with highly
conducting contacts (e.g., the source and drain contacts in the case of a field-effect
transistor). The edges of the 2DEG channel result in the reflection of the propagating
plasma waves and the formation of the standing plasma waves (plasma oscillations).
Due to the plasma wave the spectrum is quantized. If the potential at the highly
conducting contacts separated by the spacing L is fixed (so that, e.g.,
jw x = 0 = jw x = L = 0 ), a simplified quantization rule yields q = qn = pn / L , where n
= 1, 2, 3, … is the index of the oscillations mode. In particular, Equation (8.42) for
a HEMT structure with the gate length Lg » L (i.e., Lc< < Lg) yields w » nWg,
where:
4pe2 ∑ 0 W
Wg =
kmL2g (8.43)
is the fundamental plasma frequency in the gated 2DEG channel. By applying

boundary conditions corresponding to a highly conducting contact at x = 0
( jw x = 0 = 0 ) and to a free 2DEG channel edge at x = L ( djw / dx x = L = 0 ), the
wavenumbers are qn = pn / 2 L . More detailed and careful consideration shows that
actually the frequencies of plasma oscillations depend on the shape and conduc-
tivity of the contacts (Ryzhii et al. 2003, 2004 a-b; Satou et al. 2005 a-b) as well
as the properties of the substrate (Satou et al. ). Although the effect of boundary
conditions on plasma frequencies is not so strong and, therefore, can be neglected
in many practical applications, the boundary conditions can be essential for the
damping and growth (instability) of the plasma oscillations (Dyakonov and Shur
1993; Crowne 1997; Ryzhii et al. 2005).
The quantization rule for the partially gated 2DEG channel (this is common for
many field-effect transistor structures in which the channel sections adjacent to the
source and drain are ungated) is also complicated by the channel nonuniformity.
The ungated sections of the 2DEG channel do not significantly affect the spectrum
of plasma oscillations for lengths Lc smaller than Lg (Satou et al. 2003). However,
when Lc becomes comparable with Lg, the fundamental plasma frequency decreases
markedly with increasing Lc. The spectra of plasma oscillations in the 2DEG chan-
nels with a periodic system of highly conducting gates (metal grating) depend on
the period Lc+ Lg and the ratio Lc/Lg; see, for example, Krasheninnikov and Chaplik
(1981) and Matov et al. (1998).
The plasma oscillation frequencies in the 2DEG channels of the real heterostruc-
ture devices fall into the THz range. Indeed, assuming for a gated GaAs 2DEG
channel that ∑ = 1 × 1012 cm −2 , Lg ≈ L = 1 × 10 −4 cm , and W = 1 × 10−5, for the fun-
damental plasma frequency, one obtains Wg / 2p ≈ THz . Because ∑ 0 in the gated
0
channels depends on the gate voltage Vg, the plasma frequencies in such channels
can be tuned by this voltage. Fig. 8.11 shows the variation of the fundamental plasma
frequency Wg / 2p and the quality factor Q = 4Wg / pn with changing gate voltage
Vg calculated for HEMT structures with relatively short ungated gate-source and
drain-gate regions (Lc< < Lg).
1.25 120
ν = 2x1011s−1 Lg = 0.5 μm
Plasma frequency, THz 1.00 = 0.75 μm
= 1.0 μm 90
Quality factor
0.75
60
0.50
30
0.25
0.00 0
0 0.2 0.4 0.6 0.8 1
Gate voltage swing, V
Fig. 8.11 Fundamental plasma frequency and quality factor versus applied gate voltage calcu-
lated for HEMT structures with different gate length (Lg> Lc)
Owing to the nonlinearity of the hydrodynamic equations governing the plasma

waves and oscillations, different nonlinear plasma effects in 2DEG channels can be
essential and used for practical applications (Dyakonov and Shur 1996; Rudin et al.
1999; Govorov et al. 1999; Cheremisin and Samsonidze 1999).
Resonant Detection of THz Radiation Using Excitation

of Plasma Oscillations
Plasma effects in HEMT structures such as that shown in Fig. 8.10 can be used for
resonant detection of THz radiation. As proposed previously (Dyakonov and Shur
1996), the excitation of plasma oscillations by incoming THz radiation in the
HEMT 2DEG channel results in variations of the dc current in the channel or of the
dc voltage between the source and the drain. This is due to a rectification effect
associated with hydrodynamic nonlinearity of Equations (8.1) and (8.2) governing
the plasma oscillations. The observation of the resonant detection of THz radiation
in HEMTs was reported in several publications (Lu et al. 1998; Lu and Shur 2001;
Deng et al. 2002; Teppe et al. 2005). THz detectors based on HEMT-like structures
utilizing the excitation of plasma oscillations associated with different mechanisms
of nonlinearity were proposed in Satou et al. (2003), Khmyrova and Ryzhii (2000),
and Ryzhii et al. (2000 a-b).
In these devices, the rectified component of the dc current through the gate
structure with a nonlinear dependence of the leakage channel-gate current is used
as the measurand. The described nonlinearity can be associated with strong
(exponential) nonlinear dependences of the leakage current on the local potential in

the 2DEG channel due to thermionic, tunneling, or resonant-tunneling origin of this
current. Because the nonlinearities associated with thermionic, tunneling, and reso-
nant tunneling mechanisms are fairly strong, the responsivity of the detectors using
former mechanisms can significantly exceed the responsivity of the resonant detec-
tors using hydrodynamic nonlinearities. The resonant detectors of THz radiation
utilizing the nonlinearity of the electron transverse transport can also be based on
hot-electron transistors with both thermionic and tunneling injection (Ryzhii 1997,
1998). Recently, THz detectors based on an ungated or gated 2DEG channel with
a lateral Schottky junction (which provides a nonlinearity necessary for the rectifi-
cation) has been proposed (see Fig. 8.12). As shown, the ratio of the responsivity
Rw of the resonant detector based on a HEMT-like structure with a gated channel
(in which the plasma oscillations are excited) shown in Fig. 8.12b to the responsiv-
ity R0 of the usual Schottky junction detector (without the excitation of plasma
oscillations) can be expressed by
Rw 1
≈ (8.44)
R0 ⎡⎣sinh (pn / 4Wg ) + cos2 (pw / 2Wg )⎤⎦
2
(a) V0+Vω cos ωt
+ + + + + + + + + + + + +
− − − − − − − − − − − − −
Ohmic Quasi-neutral Depletion Schottky

contact 2DEG channel region contact
(b) V0+Vω cos ωt

Gate
+ + + + + + + + + + + + +
− − − − − − − − − − − − −
L l
(c)
eVbi
eV0
Fig. 8.12 Structures of THz detectors based on lateral Schottky junction with (a) an ungated, (b)
gated channel, and (c) device band diagram under forward bias (V0 < 0)
μ = 30x104cm2 /Vs
= 12x104cm2 /Vs
1000 = 6x104cm2 /Vs
Responsivity, a.u. 100
10
1
0.5 1 1.5 2 2.5
Frequency, THz
Fig. 8.13 Normalized responsivity as function of THz radiation frequency calculated for detec-
tors based on lateral Schottky junction with different electron mobility in the 2DEG channel
The same ratio but for a resonant detector with an ungated channel (see Fig.
8.12a) is given
Rw 1
≈ (8.45)
R0 ⎡⎣sinh (pnw / 4Wu ) + cos2 (pw 2 / 2Wu2 )⎤⎦
2 2
where (compare with Equation (8.43))
p 2 e2 ∑ 0
Wu = (8.46)
kmL
is the fundamental plasma frequency in the ungated 2DEG channel, as shown in

Fig. 8.13.
Due to the large amplitude of the plasma oscillations forced by the incoming THz
signal, the nonlinearities in question lead not only to the occurrence of the rectified
component of the terminal current (or the pertinent voltage) but to the occurrence of
higher harmonics. Hence, the devices under discussion in this section can be used
for plasma-assisted resonant detection as well as frequency multiplication.
Comments
The concepts discussed above can result in the development of novel THz heter-
ostructure devices such as detectors and frequency multipliers of THz radiation.
However, the device proposals considered do not exhaust all interesting new ideas;
see, for example, Govorov et al. (1998, 1999), Ryzhii et al. (2002 a-b), and
Hanabe et al. (2005). The specific features of plasma waves in the 2DEG channel
can be used not only in the devices like those discussed above. The linearity of
plasma wave dispersion ( Re w α q ), relatively large plasma wave phase and group
velocity (S » 108cm/s), and the possibility to control the plasma wave propagation
and the interaction plasma waves between each other by locally applied voltage
open up wide prospects to create new systems for processing of THz signals (in
particular, delay lines, interferometers, etc.).
Ultrasensitive THz Detector Using Cold-Electron Bolometer
Cosmology experiments in the last few years (BOOMERanG, WMAP) have dis-
covered that the universe consists of 73% dark energy, 23% dark matter, and only
4% ordinary matter. The most shocking news is the acceleration of the universe by
unknown forces (Breakthrough of the Year: Illuminating the Dark Universe 2003).
Experiments to resolve the nature of these mysterious dark components will require
a new generation of ultrasensitive detectors to get a more detailed picture of the
cosmic microwave background radiation (Breakthrough of the Year: Illuminating
the Dark Universe 2003).
A new principle to realize an ultrasensitive THz detector was proposed by Kuzmin
et al. (Kuzmin 2000; Kuzmin et al. 1998; Kuzmin and Golubev 2002). A novel con-
cept of the cold-electron bolometer (CEB) is based on strong direct electron cooling
of the absorber removing all incoming power from the supersensitive absorber to the
readout system with considerably higher dynamic range. This concept is purposed to
overcome the main contradiction of supersensitive detectors: overheating by back-
ground power load due to high sensitivity of the detector. Moreover, additional artifi-
cial dc heating of the TES (transition-edge sensor) for electrothermal feedback will
be replaced by effective electron cooling (see Fig. 8.14). This could entail a
Cold-Electron Bolometer (CEB) Transition-Edge Sensor (TES)

P0 -removed by SIN junctions Ptotal = P0 + Pbias, Pbias = Pmax signal
time
time
electron cooling! Pbias- heating!

2 2
Te cool Te heat
(a) (b)
P0 P0
0 0
100 mk 230 mk Te 100 mk 230 mk Te
Tph Tph
Fig. 8.14 Comparison of the CEB and TES concepts (Kuzmin 2004)
significant breakthrough in the development of supersensitive THz detectors due to

the following.
In both concepts the background power load leads to overheating of the small
absorber to the same temperature for a comparable volume of absorbers (V = 0.05 mm3).
However, in turning point “2”, the CEB can cool down the electron temperature
T back to phonon temperatures or lower due to direct electron cooling. In contrast,
the TES needs an additional dc heating to Tc for electrothermal feedback. The
advantage of cooling in comparison with heating is evident for supersensitive
detectors.
• Achieving high sensitivity with direct electron cooling of an absorber with elec-
tron temperatures lower than bath temperature with the corresponding improve-
ment of noise properties.
• CEB creates a new opportunity to avoid saturation by removing background
power from the absorber (strong electrothermal feedback) by means of direct
electron cooling.
The CEB can be easily fabricated on planar substrates in the form of multipixel
arrays with possible multiplication using a SQUID or HEMT readout. The CEB
concept is in the process of development for a new generation of balloon-borne
telescopes OLIMPO, CLOVER, and PILOT. The goal of the first stage will be to
achieve a noise equivalent power (NEP) of the order of 10–18 W/Hz1/2 with a SQUID
readout system at 300 mK in voltage-biased mode.
The most developed superconducting bolometer (built in the 1970s) is the
transition-edge sensor. Some progress has been achieved after the introduction of
electrothermal feedback (Lee et al. 1996). Currently, the TES is the most widespread
bolometer with a SQUID readout system available for multiplexing. However, the
TES has severe problems with saturation and the most drastic problem is artificial
overheating by dc power for the feedback (after point “2” in Fig. 8.14b). Additional
heating requirements make all efforts in the area of deep cooling very challenging
and do not look promising in terms of attaining the limit performance of the
bolometer.
In contrast to this overheating, the new concept of a cold-electron bolometer with
direct electron cooling (Fig. 8.15) was introduced by Kuzmin et al. (Kuzmin 2000;
Kuzmin et al. 1998; Kuzmin and Golubev 2002). The CEB is the only active concept
suggesting the removal of incoming background power from the supersensitive
region of the absorber (point “2” in Fig. 8.14a). This concept is likely to prevail in the
long run over concepts requiring heating of the TES because it returns the system to
the lowest temperature (noise) state. In this state, the system shows the most
responsivity to incoming THz signals and improved noise properties. All the power
of the signal is detected in measurements. This bolometer can be especially effective
for operation in the presence of a realistic background power load.
Theoretical estimations and preliminary experiments show that it is possible to
realize the necessary sensitivity of better than 10−18 W/Hz1/2 with an antenna-
coupled CEB at a temperature of £0.3 K (Kuzmin and Golubev 2002). Additional
advantages of such detectors include the possibility to operate in a wide range of
Fig. 8.15 Capacitively coupled CEB with SIN tunnel junctions for temperature measurements
and electron cooling (strong electrothermal feedback) [Kuzmin 2004]. An AFM picture of the
right part of the bolometer shows SIN junction and a part of the large antenna
Fig. 8.16 Record electron cooling achieved at Chalmers University in real bolometer configuration
due to improved quasiparticle trapping: (a) Au trap just near the junctions; (b) improved shape of
superconducting electrode; (c) usual cross geometry [Kuzmin et al. 2004, Agulo et al. 2004]
background load, easy integration in arrays, and the possibility of polarization

measurements.
The effect of nonequilibrium electron cooling of CEB has been demonstrated by
Nahum et al. (Kuzmin 2004) for normal metal strip connected to SIN tunnel junc-
tions. The results of the Jyväskylä group on electron cooling from 300 to 110 mK
are attracting strong interest from ESA.
The proposed NASA/ESA missions SPIRIT, SPECS, SAFIR, and “Far IR Proto-
Galaxy Imager” will determine the highest level of requirements for bolometers in the
nearest future. No existing technology could satisfy these requirements. Technological
breakthrough is needed, first of all, to approach these requirements. The proposed
CEB concept could be a good candidate to become a leading concept in this
development.
The latest achievement of the Chalmers group is a record electron cooling from
290 to 93 mK (Fig. 8.16) due to improved trapping of hot quasi-particles in a super-
conductor (Masi et al. 2004; Kuzmin and Mauskopf 2005). The achieved cooling
results give a good basis for realization of high-performance CEBs working at real
conditions of background power load.
Optimization of the CEB in Presence of the Background

Power Load
Model: Here we assume that the SIN tunnel junctions are voltage-biased, and the
current is measured by SQUID. The sensitivity of the device is then characterized
by the current responsivity SI, which is the ratio of the current change detected by
the SQUID and the change in the power load of the bolometer caused by a detected
signal:
∂I ∂I∂T
SI = w = .
∂Pw ∂P (8.47)
−i w cv L + 5SLTe +4
∂T
Here cn = gTe is the specific heat capacity of the normal metal; 5 S L Te4 is the
thermal conductance between electron to the phonon subsystems in the normal
metal, ∑ is a material constant, ∧ is a volume of the absorber, Te and Tph are the
electron and phonon temperatures of the absorber; ∂P / ∂T is the thermal conduct-
ance of the NIS junction, and P(t) is the incoming radio frequency power.
The noise is captured by the noise equivalent power, which is the sum of three
different contributions, and is defined as
dI 2
2
NEPtotal = NEPe2− ph + NEPSIN
2
+ . (8.48)
SI2
Here,
NEPe2− ph = 10 kB SL (Te6 + Tph6 ) (8.49)
is the nonequilibrium noise associated with electron–phonon interaction; NEP2SIN is

2
the noise of the SIN tunnel junctions, and the last term dI is the noise of an
2
SI
amplifier (SQUID), dI , which is expressed in pA/Hz1/2 (Kuzmin et al. 2004; Agulo
et al. 2004).
The noise of the NIS tunnel junctions, NEP2SIN, has three components: shot noise
2eI/S2I, the fluctuations of the heat flow through the tunnel junctions, and the cor-
relation term between these two processes:
dPw dIw dIw2

2
NEPSIN = dPw2 − 2 + 2 . (8.50)
SI SI
−18
8x10
NEP
Goal: NEP=10−18 W
6 Po= 0, R=6 k Ω
10−13 W, 6 k Ω
NEP (W/Hz1/2 )
10−13 W, 0,5 k Ω
4 Po- microwave background load
1/2
2 δ ISQUID= 50 fA/Hz
0
0.75 0.80 0.85 0.90 0.95 1.00 1.05
V/Δ
Fig. 8.17 NEP in presence of various background power loads and various efficiencies of direct
electronic cooling for bath temperature 100 mK (Kuzmin and Golubev 2002)
It is necessary to take into account the effect of the electron cooling of the metallic
strip by the NIS tunnel junctions.
Effect of background power load: Our analysis of the effect of background
power load on noise performance for different configurations of CEB bolometers
shows that the optimal configuration of the bolometer is a CEB with voltage-biased
SIN tunnel junctions and a SQUID readout (Kuzmin et al. 2004). The volume of
the absorber is equal to 0.05 mm3, which is typical for our experiments.
The current noise of SQUID is equal to 50 fA/Hz1/2 in our simulations. The
results are shown in Fig. 8.17 for two levels of microwave background power: P0=
0 and 0.1 pW. The latter figure is a realistic background power load P0 for a band-
width of 10% at frequencies in the range of 300–1000 GHz for background tem-
perature Tbg= 3 K. The first curve without background load (P0= 0) produces NEP
= 2 × 10−19 W for typical junction resistance (R) equal to 6 kW.
A considerable increase of the NEP to 8 × 10−18 W/Hz1/2 is obtained for P0= 0.1
pW. The electron temperature also increases from 100 to 230 mK. Decreasing R to
0.5 kW improves the efficiency of the electronic cooling and returns the NEP to the
acceptable level of 8 × 10−19 W/Hz1/2 and Te to the level of 100 mK. The NEP goal
for the future projects is 10−18 W/Hz1/2 (Kuzmin 2000; Kuzmin et al. 1998; Kuzmin
and Golubev 2002) and can be achieved with these system parameters.
Concept of an optimal bolometer: We have analyzed the optimal CEB in the
presence of the final background power load (P0 = 0.1 pW) for fixed parameters of
the SQUID-amplifier (10 fA/Hz1/2) at T = 300 mK (Agulo et al. 2004). The optimal
regime can be realized when thermal “cooling conductance” through the tunnel
junctions dominates the “fundamental” electron–phonon conductance. In these
circumstances, an NEP level of 10−18 W/Hz1/2 at 300 mK can be achieved.
The dependences of the NEP on a volume of the absorber show that there is no
optimal value of NEP for the volume of absorber L. The reason for the flattening
for small volumes is that we have achieved full transference of P0 to the amplifier,
so that the NEPe-ph constitutes less than 50% of the total NEP. The critical point of
the optimal regime is a point of equality of NEPSIN and NEPe-ph at L = 0.003 mm3.
The dependences of the NEP on the resistance of the SIN tunnel junctions R gives
the optimum value R around 1.5 kW. For higher values of R, the electron cooling is
not as effective and responsivity is decreased, increasing noise of the SIN junction
(4) and SQUID (2). Decrease of R lower than the optimal point increases the shot
noise (reverse proportional to R) without any increase in responsivity because of
saturation in transferring power.
Expected results: Improvement of noise equivalent power of the bolometer
receiver due to realization of the optimal concept to the level of NEP ~ 10−18
W/Hz1/2 at 300 mK is expected.
Ultimate Noise Performance of CEB-General NEP Formula
This question has arisen in relation to the highest requirements on NEP for future
NASA missions. The question is how realistic are these requirements on NEP =
10−20 W/Hz1/2.
The ultimate performance of CEB and other concepts has been analyzed
(Kuzmin 2004). The NEP is determined by shot noise due to power load. Other
sources of noise are neglected due to small values. For the level of P0 = 10 fW this
limit can be achieved using relatively low temperatures (~100 mK) and small volume
of the absorber (∧ ≤ 0.002mm3) when we can neglect the electron–phonon noise
component.
A general ultimate NEP formula has been derived (Kuzmin 2004):
NEPshot = (2 P0 Equant )1/ 2 (8.51)
where P0 – background power load

Equant – energy level of P0quantization
Equant = kBTe – for normal metal absorber
Equant = Δ – for superconducting absorber
The ultimate NEP can be estimated for different bolometers for rather low P0 =
10 fW:
CEB: Te = 50 mK, NEPshot = 1*10−19 W/Hz1/2
TES: Te= 500 mK, NEPshot = 4*10−19 W/Hz1/2
KID: T = 2 K (Δ= 200 meV), NEPshot = 7*10−19 W/Hz1/2.
The lowest NEP can be achieved for CEB with the lowest level of quantization.
However, even these extreme parameters of P0 and Equant show that it is rather unre-
alistic to achieve NEP = 10−20 W/Hz1/2 as announced in NASA requirements for
future missions.
This formula (8.51) is used for estimation of ultimate parameters of CEB and
other bolometers for given parameters of detector systems. The work on analysis of
ultimate parameters of CEB will be prolonged.
Preliminary results: Record attowatt sensitivity of the cold-electron bolometer.
The Chalmers group has made the measurements of the cold-electron bolometer
in a current-biased mode. They have measured the voltage response to applied
low-frequency modulation of the heating current on the normal metal absorber.
The detector responsivity was 1.5 × 1010 V/W at 35 Hz at 100 mK. The frequency
dependence of responsivity was tested and was found to decrease with increasing
frequency of the modulated signal owing to the inclusion of the signal attenuation
due to cryogenic filters. The corresponding noise equivalent power of the bolo-
meter was obtained, using an operational amplifier at 300 K from the noise of the
bolometer divided by the detector responsivity (Fig. 8.18).
The record noise equivalent power for the CEB was found to be better than 10−18
W/Hz1/2 at 100 mK and at frequencies higher than 100 Hz for a background power
load of 2.6 fW. The next step is measurement of NEP in voltage-biased mode.
For the next breakthrough in our knowledge about dark matter and dark energy,
we need a new generation of detectors. Priority of this topic can be determined as
96% (dark universe) to 4% (ordinary matter).
A cold-electron bolometer is a good candidate to become the leading concept in
this development.
6
NEP Total
5 NEP Bolo
NEP Amp
NEP (10−18 w/Hz1/2)
100 1000
frequency (Hz)
Fig. 8.18 The dependence of the total, bolometer, and amplifier NEP of the cold-electron bolom-
eter to the modulation frequency. We have stepped to the 19th power of CEB sensitivity for fre-
quencies higher than 100 Hz. The dashed line represents the value of the NEPs as predicted by the
nonequilibrium theory of the CEB
Realization of the cold-electron bolometer would be the turning point from arti-
ficial heating (TES) to effective electron cooling lower than phonon temperature
which should bring clear benefits for supersensitive detectors.
The famous contradiction between supersensitivity and saturation can be over-
come by strong electrothermal feedback removing power to the next stage with
higher dynamic range.
The CEB concept could be implemented for new balloon telescopes OLIMPO,
CLOVER, and PILOT.
Summary
We have discussed sensing of two parts of the EM spectra, namely the IR region
and the THz region. For the IR region we have particularly analyzed QWIP, QRIP,
and QDIP and found that QRIP and QDIP should be superior to QWIP for IR detection.
For the THz region we first analyzed how influence of strain on impurities can be
used for THz detection (and generation). Then we analyzed in detail plasma effects
in two-dimensional electron systems for sensing (and generation). Finally we gave
an example of an ultrasensitive THz detector for space application which has shown
excellent experimental figures of merit.
References
Agulo I, Kuzmin L, et al. (2004) Nanotechnology, 15:224–228

Allen SJ, Tsui DC, Logan RA (1977) Phys. Rev. Lett., 38:980–983
Altukhov IV, Chirikova EG, Kagan MS, Korolev KA, Sinis VP, Smirnov FA (1992) Sov. Phys.
JETP, 74:404
Andronov AA (1987) Sov. Phys. Semiconduct., 21:701
Bannov NA, Ryzhii VI (1983) Sov. Phys. Semicond., 17:439–441
Berryman KW, Lyon SA, Segev M (1997) Appl. Phys. Lett., 70:1861–1863
Bhattacharya P, Stiff-Roberts AD, Krishna S, Kennerly S (2002) Proc. SPIE, 4646:100–106
Boucaud P, Brunhes T, Sauvage S, Yam N, Thanh V Le, Boucier D, Rappaport N (2001) Phys.
Stat. Sol. (a), 224:233–235
Brazis R (1995) Lithuanian Phys. J., 35:447
Breakthrough of the Year: Illuminating the Dark Universe (2003) Science, 302, 2038
Chaplik AV (1972) Sov. Phys. JETP, 35:395–398
Cheremisin MV, Samsonidze GG (1999) Semiconductors, 33:578–582
Choi KK (1997) The Physics of Quantum Well Infrared Photodetectors. Singapore: World
Scientific
Crowne J (1997) J. Appl. Phys., 82:1242–1254
Deng Y, Knap W, Rumyantsev S, Gaska R, Khan A, Ryzhii V, Kaminska E, Piotrowska A,
Lusakowski J, Shur MS (2002) 2002 IEEE Lester Eastman Conference on High Performance
Devices, University of Delaware, Newark, pp. 135–142
Duboz J-Y, Liu HC, Wasilewski ZR, Byloss M, Dudek R (2003) J. Appl. Phys., 93:1320–1322
Dyakonov M, Shur M (1993) Phys. Rev. Lett., 71:2465–2468
Dyakonov M, Shur M (1996) IEEE Trans. Electron. Dev., 43:380–387, 1640–1645
Einevoll GT, Chang Y-C (1990) Phys. Rev. B, 41,1447

Ekenstedt MJ, Andersson TG, Wang SM (1993) Phys. Rev. B, 48:5289
Ershov M, Ryzhii V, Hamaguchi C (1995) Appl. Phys. Lett., 67:3147–3149
Faist J, Capasso F, Sivco DL, Hutchinson AL, Cho AY (1994) Science, 264:553
Fraizzoli S, Pasquarello A (1990) Phys. Rev. B, 42:5349
Fraizzoli S, Pasquarello A (1991) Phys. Rev. B, 44:1118
Fritz IJ (1987) Appl. Phys. Lett., 51:1080
Govorov AO, Kovalev VM, Chaplik AV (1999) JEPT Lett., 70:488–490
Govorov AO, Studenikin SA, Frank WR (1998) Phys. Rev. B, 58:1517–1532
Grave I, Yariv A (1992) In: Intersubband Transitions in Quantum Wells (Rosencher E, Vinter B,
Levine B eds.). New York: Plenum, p. 15
Hanabe M, Otsuji T, Ishibashi T, Uno T, Ryzhii V (2005) Jpn. J. Appl. Phys., 44:3842–3847
Helm M, Hilber W, Fromherz T, Peters FM, Alavi K, Psthar RN (1994). In: Quantum Well
Intersubband Transition Physics and Devices (Liu HC, Levine BF, Andersson JY, eds.).
Dordrecht: Kluwer, p. 291
Holtz PO, Zhao QX (2005) In: Materials Science V77, Impurities Confined in Quantum Structures
(Hull R, Parisi J, Osgood RM, Warlimont H, eds.). Springer
Horiguchi N, Futatsugi T, Nakata Y, Yokoyama N, Mankad T, Petroff PM (1999) Jpn. J. Appl.
Phys., 38:2559–2561
Hu BYK, Wilkins JW (1991) Phys. Rev. B, 43:14009–14029
Kempa K, Bakshi P, Xie H (1993) Phys. Rev. B, 48:9158–9161
Kersting R, Unterrainer K, Strasser G, Kauffmann HF, Gornik E (1997) Phys. Rev. Lett.,
79:3038–3041
Khmyrova I, Ryzhii V (2000) Jpn. J. Appl. Phys., 39:4727–4732
Kim S, Mohseni H, Erdtmann M, Michel E, Jelen C, Razeghi M (1998) Appl. Phys. Lett.,
73:963–965
Kochman B, Stiff-Roberts D, Chakrabarti S, Phillips JD, Krishna S, Bhattacharya P (2003) IEEE
J. Quant. Electron., 39:459–463
Krasheninnikov MV, Chaplik AV (1981) Sov. Phys. Semicond., 15:19–22
Kustov VL, Ryzhii VI, Sigov YUS (1980) Sov. Phys. JETP, 52:1207–1212
Kuzmin L (2000) Proceedings of the 9th Symposium on Space THz Technology, Pasadena, pp.
99–103, March 1998; Phys. B: Condensed Matt., 284–288, 2129
Kuzmin L (2004) Proceedings of SPIE conference “Mm and Submm Detectors”, Glasgow, June
21–25, Vol. 5498, pp. 349–361
Kuzmin L, Golubev D (2002) Phys. C, 372–376:378–382
Kuzmin L, Mauskopf P (2005) Superconducting Cold-Electron Bolometer with JFET Readout for
Balloon-Borne Telescope OLIMPO, ISSTT-05, Gothenburg, May
Kuzmin L, Devyatov I, Golubev D (1998) Proceedings of SPIE, 3465:193–199
Kuzmin L, Agulo I, et al. (2004) Supercond. Sci. Technol., 17:400–405
Lee A, Richards P, Nam S, Cabrera B, Irwin K (1996) Appl. Phys. Lett., 69:1801
Lee S-W, Hirakawa K, Shimada Y (1999) Appl. Phys. Lett., 75:1428–1430
Li L-X, Sun S, Chang Y-C (2003) Infrared Phys. Technol., 44:57–61
Liu HC (1992) Appl. Phys. Lett., 60:1507–1509
Liu HC, Gao M, McCaffrey J, Wasilewski ZR, Fafard S (2001) Appl. Phys. Lett., 78:79–81
Liu HC, Gunapala SD, Schneider H (eds.) (2004) QWIP 2004: Proceedings of the International
Workshop on Quantum Well Infrared Photodetectors, Elsevier, Amsterdam
Loehr JP, Chen YC, Biswas D, Bhattacharya PK, Singh J (1990) Proceedings of the 20th
International Conference on the Physics of Semiconductors, World Scientific, Singapore, Vol.
2, p. 1404
Lu J-Q, Shur MS (2001) Appl. Phys. Lett., 78:2587–2589
Lu J-Q, Shur MS, Hesler JL, Sun L, Weikle R (1998) IEEE Electron. Dev. Lett., 19:373–374
Luryi S (1985) IEEE Electron. Dev. Lett., 6:78–80
Maimon S, Finkman E, Bahir G, Schacham SE, Garcia JM, Petroff PM (1998) Appl. Phys. Lett.,
73:2003–2005
Masi S, de Bernardis P, et al. (2006) In: Background Microwave Radiation and Intracluster
Cosmology eds, Melchiorri F and Rephaeli Y (New IOS press publication)
Masselink WT, Change Y-C, Morkoc H (1983) Phys. Rev. B, 28:7373
Masselink WT, Change Y-C, Morkoc H (1985) Phys. Rev. B, 32:5190
Matov OR, Meskov OF, Popov VV (1998) J. Exp. Theor. Phys., 86:538–544
Matthews JW, Blakeslee AE (1974) J. Crystal Growth, 27:118
Miesner C, Brunner K, Abstreiter G (2001) Phys. Stat. Sol. (a), 224:605–607
Nakayama M (1974) J. Phys. Soc. Jpn., 36:393–398
Odnoblyudov MA, Chistyakov VM, Yassievich IN, Kagan MS (1998) Phys. Stat. Sol. (b),
210:873
Odnoblyudov MA, Yassievich IN, Kagan MS, Galperin YUM, Chao KA (1999) Phys. Rev. Lett.,
83:644
Pan D, Towe E, Kennerly S (1998) Appl. Phys. Lett., 73:1937–1939
Pan D, Towe E, Kennerly S (1999) Appl. Phys. Lett., 75:2719–2721
Pasquarello A, Andreani LC, Buczko R (1989) Phys. Rev. B, 40:5602
People R, Bean JC (1985) Phys. Appl. Lett., 47:322
Phillips J, Kamath K, Bhattacharya B (1998) Appl. Phys. Lett., 72:2020–2022
Phillips J, Bhattacharya P, Kennerly SW, Beekman DW, Dutta M (1999) IEEE J. Quant. Electron.,
35:936–940
Raghavan S, Rotella P, Stinz A, Fuchs B, Krishna S, Morath C, Cardimona DA, Kennerly SW
(2002) Appl. Phys. Lett., 81:1369–1371
Rappaport N, Finkman E, Brunhes T, Boucaud P, Sauvage S, Yam N, Thanh V Le, Boucier D
(2000) Appl. Phys. Lett., 77:3224–3226
Rokhinson LP,Chen CJ, Tsui DC, Vawter GA, Choi KK (1999) Appl. Phys. Lett., 74:759–761
Rosencher E, Vinter B, Luc F, Thibaudeau L, Bois P, Nagle J (1994) IEEE Trans. Quant. Electron.,
30:2875–2882
Rudin S, Samsonidze G (1998) Phys. Rev. B, 58:16369–16373
Rudin S, Samsonidze G, Crowne F (1999) J. Appl. Phys., 86:2083–2088
Ryzhii M, Ryzhii V (1999) Jpn. J. Appl. Phys., 38:5922–5927
Ryzhii M, Ryzhii V, Willander M (1999) Jpn. J. Appl. Phys., 38:6650–6653
Ryzhii V (1996) Semicond. Sci. Technol., 11:759–765
Ryzhii V (1997) Appl. Phys. Lett., 70:2532–2534 and J. Appl. Phys., 81:6442–6448
Ryzhii V (1998) Jpn. J. Appl. Phys., 37:5937–5944
Ryzhii V (2001) Jpn. J. Appl. Phys., 40:L148–L150 and Appl. Phys. Lett., 78:3346–3348
Ryzhii V (ed.) (2003) Intersubband Infrared Photodetectors. Singapore: World Scientific
Ryzhii V, Ershov M (1995) Semicond. Sci. Technol., 10:687–690
Ryzhii VI, Fedirko VA (1983) Sov. Phys. Semicond. 17:850–851
Ryzhii V, Liu HC (1999) Jpn. J. Appl. Phys., 38:5815–5822
Ryzhii VI, Bannov NA, Fedirko VA (1984) Sov. Phys. Semicond., 18:481–493
Ryzhii V, Khmyrova I, Ryzhii M, Ershov M (1996) J. Phys. IV, 6:C3-157–159
Ryzhii V, Khmyrova I, Shur M (2000a) J. Appl. Phys., 88:2868–2871
Ryzhii V, Khmyrova I, Mitin V, Stroscio M, Willander M (2001a) Appl. Phys. Lett.,
78:3523–3525
Ryzhii V, Khmyrova I, Pipa V, Mitin V, Willander M (2001b) Semicond. Sci. Technol.,
16:331–338
Ryzhii V, Khmyrova I, Ryzhii M, Mitin V (2004a) Semicond. Sci Technol., 19:8–16
Ryzhii V, Khmyrova I, Ryzhii M, Pipa V, Mitin V, Willander M (2001c ) Proc. SPIE,
4288:396–406
Ryzhii V, Khmyrova I, Shur M (2002a), J. Appl. Phys., 91:1875–1881
Ryzhii V, Pipa V, Khmyrova I, Mitin V, Willander M (2000b) Jpn. J. Appl. Phys.,
39:L1283–L1285
Ryzhii V, Ryzhii M, Liu HC (2002b) J. Appl. Phys., 92:207–213
Ryzhii V, Satou A, Shur MS (2003) J. Appl. Phys., 93:10041–10045
Ryzhii V, Satou A, Shur MS (2005) Phys. Stat. Sol. (a), 202:R113–R115
Ryzhii V, Satou V, Khmyrova I, Chaplik A, Shur MS (2004b) J. Appl. Phys., 96:7625

Sa’ar A, Mermelstein C, Schneider H, Schoenbein C, Walther M (1998) IEEE Photon. Technol.
Lett., 10:1470–1472
Satou A, Khmyrova I, Ryzhii V, Shur MS (2003) Semicond. Sci. Technol., 18:460–469
Satou A, Ryzhii V, Chaplik A (2005a) J. Appl. Phys., 98:034502 1–5
Satou A, Vyurkov V, Khmyrova I (2004) Jpn. J. Appl. Phys., 43:L566–L568
Satou A, Khmyrova I, Chaplik A, Ryzhii V, Shur MS (2005b) Jpn. J. Appl. Phys.,
44:2592–2596
Sergeev A, Mitin V, Stroscio M (2002) Phys. B, 316–317:369–371
Shur MS (1990) Physics of Semiconductor Devices. Englewood Cliffs, NJ: Prentice-Hall
Shur MS, Ryzhii V (2003) Int. J. High Speed Electron. Syst., 13:575–600
Sirmain G, Reichertz LA, Dubon OD, Haller EE, Hansen WL, Brundermann E, Linhart AM,
Roser HP (1997) Appl. Phys. Lett., 70:1659
Stern F (1967) Phys. Rev. Lett., 18:546–548
Tang S-F, Lin S-Y, Lee S-C (2002) IEEE Trans. Electron. Dev., 49:1341–1346
Teppe F, Veksler D, Kacharovskii VYU, Dmitriev AP, Rumyantsev S, Knap W, Shur MS (2005)
Appl. Phys. Lett., 87:022102–022104
Towe E, Pan D (2000) IEEE J. Sel. Top. Quant. Electron., 6:408–413
Tsao JY, Dodson BW, Picraux ST, Cornelison DM (1987) Phys. Rev. Lett., 59:2455
Vasanelli A, De Giorgi M, Ferreira R, Cingolani R, Bastard G (2001) Phys. E, 11:41–43
Wang SM, Andersson TG, Vladimir KD, Yao JY (1991) Superlattice Microstruct., 9:123
Xu SJ, Chua SJ, Mei T, Wang XC, Zhang XH, Karunasiri G, Fan WJ, Wang CH, Jiang J, Wang S,
Xie XG (1998) Appl. Phys. Lett., 73:3153–3155
Yakimov AI, Dvurechenskii AV, Proskuryakov YYU, Nikiforov AI, Pchelyakov OP, Teys SA,
Gutakovskii AK (1999) Appl. Phys. Lett., 75:1413–1415
Ye Z, Campbell JC, Chen Z, Kim E-T, Madhukar A (2002) IEEE J. Quant. Electron.,
38:1234–1239
Zhao QX, Willander M (1998) Phys. Rev. B, 57:13033
Zhao QX, Willander M (1999) J. Appl. Phys., 86:5624
Zhao QX, Willander M (2000) Phys. Lett. A, 270:273
Zhao QX, Holtz PO, Pasquarello A, Monemar B, Willander M (1994a) Phys. Rev. B, 50:2393
Zhao QX, Karlsteen M, Willander M, Wang SM, Sadeghi M (2000) Phys. Rev. B, 62:5055
Zhao QX, Pasquarello A, Holtz PO, Monemar B, Willander M (1994b) Phys. Rev. B, 50:10953
Zhao QX, Wongmanerod S, Willander M, Holtz PO, Wang SM, Sadeghi M (2001) Phys. Rev. B,
19:5317
Index
A D
Absorption Deposition 42, 45, 53, 54, 55, 58, 59, 60, 61,
Optical 14, 15 62, 63, 70, 76, 102, 110, 127, 128, 132
Actuator 2 Glancing-angle 32, 36, 48, 63, 64, 77, 78
Accuracy 2, 5, 8, 23 Oblique angle 46, 47, 48, 50, 63
Aging 3, 8 Material 2
Architecture Process 21, 31
Sensor 3, 4 Pulsed-laser 36
Attributes Vapor 31, 32, 33, 35, 36, 37, 45, 77,
Performance 4, 5 78, 149, 152, 187
Electrochemical 32, 41, 63, 69, 77, 78
Thin film 32, 38, 46, 70, 76, 93, 100,
B 102, 112, 115
Behavior Nanostructure 77, 78
plastic 12, 13 Sol-gel 151, 152
elastic 12 Detector
Photo 94, 133, 170, 171, 172, 173, 174,
181, 183, 186, 208, 209, 210
C Dielectric 10, 15, 16
Calibration 3, 6, 8, 9, 99, 106, 107, 120, Diffusion 7, 8, 21, 100, 127, 140, 153
121, 122, 126, 133, 164 Drift 5, 8, 28, 94, 101, 130, 132, 138
Ceramic 11, 14, 39, 70, 81, 149
Conductivity
Thermal 11, 18, 93 E
Electrical 18, 82, 111 Electronics 3, 5, 7, 9, 29, 80, 81, 82, 84, 96,
Ionic 114, 118, 120, 121, 168 97, 100, 109, 134, 147, 161, 178
Photo 190, 195 Energy 3, 6, 10, 11, 12, 13, 14, 15, 17, 18,
Confinement 21, 27, 28, 36, 39, 68, 70, 71, 82, 88,
Geometrical 3, 11, 12, 14, 16 89, 90, 96, 101, 102, 106, 107, 111,
Quantum 169, 186, 187 112, 113, 114, 123, 124, 129, 137, 170,
Curve 171, 173, 174, 176, 178, 185, 188, 189,
Calibration 3, 120, 121, 122, 133 190, 191, 192, 193, 201, 206, 207
Response 5, 23, 95, 96, 97, 122, 151 Surface 11, 13
211
212 Index
Etching 2, 31, 32, 41, 42, 45, 70, 71, 72, 76, MEMS 2, 16, 19, 21, 22, 24, 29, 68, 69,
78, 80, 83, 84, 165 74, 75, 76, 79, 81, 82, 84, 87, 88,
Error 106, 129, 148, 149, 151, 152, 153,
Systematic 8 161, 167
Random 8 Metal 11, 12, 14, 15, 16, 21, 22,
36, 39, 44
Miniaturization 2, 106, 137, 138, 140, 142,
F 161, 167
Fabrication 1, 2, 9, 15, 20, 24, 31, 32, 33, 34, Modulus
36, 38, 39, 44, 51, 55, 58, 59, 62, 63, Elastic/Young’s 11, 13, 16, 140, 151, 156,
64, 65, 66, 69, 70, 71, 73, 74, 75, 77, 159
79, 80, 81, 82, 83, 84, 85, 87, 108, 114, Model 9, 12, 13, 14, 15, 16, 19, 39, 63, 132,
135, 137, 142, 144, 147, 151, 161, 162, 137, 138, 142, 144, 145, 176, 181, 190,
164, 165, 167, 171 191, 192, 193, 195, 204
Force Multilayers 2, 98
Dispersion 11, 123
London 11
Van Der Waals 11 N
Frequency Nano-
Resonance 2, 6 Structure 1, 2, 3, 7, 12, 13, 15, 16, 17, 27,
31, 32, 37, 38, 41, 44, 49, 50, 51, 52,
53, 55, 59, 60, 62, 63, 64, 72, 73, 74,
G 76, 77, 78, 79, 81, 83, 87, 129, 134,
Grains 12 137, 169, 170, 187
Technology 16, 43, 63, 64, 81, 82,
83, 85, 208
H Scale 11, 12, 13, 14, 15, 60, 61, 62, 63,
Hysteresis 5, 8, 127, 128 80, 81, 82, 85, 137, 140, 144
Homogeneity 13, 14 Materials 11, 12, 13, 14, 63, 81, 114, 129,
134
Morphology 7
I Fluid 13, 14, 16
Index Instrument 10
Refractive 11, 16, 67, 99, 133, 134, Particle 11, 15, 16, 27, 31, 32, 41, 44,
137, 150 45, 62, 76, 77, 78, 80, 81, 107, 134,
Inhomogeneity 13, 14 135, 139, 140, 144
Interactions Regime 11, 13
Ionic 11 Noise 3, 6, 7, 8, 10, 18, 24, 94, 95, 96, 110,
Interface 12, 61, 95, 133, 134, 137, 140, 111, 119, 129, 135, 137, 138, 150, 161,
172, 187 187, 202, 204, 205, 206, 207
L P
Liquid 11, 12, 13, 14, 28, 32, 34, 38, 39, Point
41, 44, 67, 75, 89, 95, 105, 109, 110, Melting 11, 12, 16
116, 120, 122, 123, 139, 167 Polymer 14, 16, 21, 22, 23, 28, 31, 39, 40,
41, 42, 43, 62, 63, 66, 70, 72, 73, 76,
77, 78, 79, 80, 82, 83, 107, 108, 111,
M 112, 114, 116, 119, 123, 124, 125,
Materials 126, 127, 129, 130, 131, 132, 134,
Bulk 11, 15 135, 138, 139, 140, 141, 142, 143,
Selection 1, 9, 16, 147 144, 145, 150, 153, 154
Measurand 3, 4, 5, 6, 7, 8, 9, 10, 17, 18, 19, Precision 2, 5, 8, 24, 29, 100, 102,
21, 22, 23, 24, 25, 26, 27, 28, 74, 198 137, 161
Index 213
Process Spectrum 3, 15, 18, 26, 89, 90, 91,

Fabrication 2, 9, 20, 32, 51, 62, 66, 74, 96, 101, 102, 103, 173, 174,
147, 161, 162, 164, 165 175, 196, 197
Properties Stress 8, 18, 19, 20, 21, 24, 27, 29, 65, 74,
Physical 132, 145 147, 148, 150, 152, 154, 157, 158,
Chemical 2, 11 159, 160, 163, 164, 166, 167,
Mechanical 11, 127 190, 193
Optical 14, 15, 16, 62, 63, 64, Surface 2, 3, 11, 12, 13, 14, 19, 21, 24,
134, 135, 139 27, 29, 35, 38, 41, 45, 46, 48, 49,
50, 55, 59, 60, 61, 68, 69, 70, 71,
72, 74, 75, 81, 83, 89, 96, 98, 102,
Q 107, 112, 113, 125, 130, 132, 134,
Quantitative 135, 138, 140,
Sensor 4, 5, 132 142, 144, 149, 159, 164, 167
System 2, 3, 7, 9, 11, 24, 28, 33, 39, 45,
46, 53, 62, 63, 67, 68, 69, 71, 80,
R 81, 82, 83, 88, 89, 93, 94, 95, 96,
Range 97, 100, 101, 103, 105, 106, 107,
Dynamic 5, 21, 22, 108, 109, 149, 109, 110, 120, 126, 127, 132, 134,
150, 201, 208 135, 137, 138, 139, 140, 142, 143,
Reliability 5, 10, 15, 108 144, 145, 149, 151, 161, 162, 167,
Repeatability 8 174, 188, 190, 193, 195, 196, 197,
Resolution 5, 7, 8, 24, 66, 67, 68, 70, 80, 201, 202, 204, 205, 207, 208
81, 84, 108, 134, 135, 137
Rheology 10
T
Temperature
S glass-transition 11, 12, 72, 116
Scaling 2, 8, 9, 10, 11, 13, 15, 16, 28, 31, Tension
32, 47, 50, 58, 69, 83, 110, 112, Surface 12, 41
132, 143, 144, 162 Thin films
Selectivity 2, 5, 8, 9, 71, 106, 108, Functional 2, 3, 13, 76, 78, 87, 148,
109, 110, 116, 117, 123, 129, 150, 161, 166
138, 165 Time
Self-assembly 2, 32, 64, 76, 77, 78, Response 5, 7, 25, 93, 94, 95, 101,
80, 106 103, 127
Semiconductor 27, 29, 41, 62, 63, 64, 65, Threshold
70, 71, 72, 74, 75, 79, 82, 93, 101, Sensor 4
114, 129, 142, 147, 150, 152, 169, Transducer 1, 2, 4, 5, 6, 7, 8, 9, 10, 14,
170, 187, 188, 193, 208, 209, 211 15, 16, 17, 18, 22, 65, 66, 67, 70,
Sensitivity 5, 6, 7, 8, 9, 13, 15, 18, 21, 73, 74, 75, 76, 77, 79, 80, 105, 106,
22, 23, 28, 60, 88, 91, 92, 93, 107, 108, 110, 111, 112, 114, 123,
94, 95, 108, 109, 112, 116, 119, 127, 129, 137, 138, 141, 142, 144,
125, 126, 127, 129, 130, 135, 145, 166
137, 143, 150, 167, 171, 182, Transduction 3, 6, 9, 10, 15, 17, 18, 19, 21,
201, 202, 204, 207, 208 22, 24, 25, 27, 28, 105, 106, 107, 109,
Signal-to-noise ratio 95, 97, 137, 138, 187 110, 113, 114, 116, 123, 129, 132,
Silicon 2, 19, 20, 21, 22, 23, 25, 26, 28, 138, 147, 148
70, 71, 72, 75, 76, 77, 80, 83, 99,
150, 152, 153, 162, 163, 164, 165,
167, 187 V
Solid 12, 13, 29, 34, 36, 39, 41, 62, 63, 72, Variability 2, 8, 97
81, 82, 83, 98, 99, 107, 116, 134, Van der Waals 11
140, 141, 142, 144, 167, 169, 192 Viscosity 11, 13, 14, 16, 18, 39, 41, 44, 195

Functional Thin Films and Nanostructures For Sensors. Synthesis, Physics, and Applications, 2009, p.224

Uploaded by

Copyright:

Available Formats

You might also like

Functional Thin Films and Nanostructures For Sensors. Synthesis, Physics, and Applications, 2009, p.224

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

Functional Thin Films and Nanostructures For Sensors. Synthesis, Physics, and Applications, 2009, p.224

Uploaded by

Copyright:

Available Formats

Functional Thin Films and

Nanostructures for Sensors

Series Editor: Dr. Radislav A. Potyrailo

For other titles published in this series, go to

Functional Thin Films and

ISBN: 978-0-387-36229-8 e-ISBN: 978-0-387-68609-7

Library of Congress Control Number: 2008944096

© Springer Science+Business Media, LLC 2009

Printed on acid-free paper

In my career I’ve found that “thinking outside the box” works

Different people think in different time frames: scientists think

Everything can be made smaller, never mind physics;

Integrated Analytical Systems

Series Editor: Dr. Radislav A. Potyrailo, GE Global Research, Niskayuna, NY

Anis Zribi is the manager of the Detection Technology Research, Development

Jeff Fortin is the manager of the Microsystems and Microfluidics Lab at GE

1 Sensor Design Guidelines ........................................................................ 1

3 Growth and Synthesis of Nanostructured Thin Films ......................... 31

4 Integrated Micromachining Technologies

5 Applications of Functional Thin Films

7 Applications of Functional Thin Films for Mechanical Sensing ......... 145

8 Sensing Infrared and Terahertz Regions by Functional Films ........... 167

Index ................................................................................................................ 211

Abstract This chapter focuses on introducing fundamental design principles of

The Big Picture

metrology and science, optimizing and controlling processes, providing security

optical computation, micromotors, chemical and biological sensors; MEMS devices

The basic function of a sensor is to selectively identify and measure a physical,

Measurand M Sensing X Y Y1 Y2 Numerical

Processor Storage Display

Fig. 1.1 Sensor architecture block diagram

Sensor Figures of Merit/Performance Attributes

Table 1.1 Sensor Performance Attributes

Input Dynamic Range

Fig. 1.2 Typical response curve of a sensor/transducer to a measurand Mi

Input Dynamic Range

Every quantitative sensor is characterized by a response curve that represents the

Although seemingly obvious, the definition of sensitivity can be confusing at times.

The overall sensitivity of the device to measurand Mi is defined by the following

Si total = Sfilter * A * Stransducer . (1.7)

t total = t transducer + t electronics . (1.9)

The second type of measurement uncertainty associated with a transducer is

Hysteresis and Drift

It is typical of a sensor designed to detect variations of a given measurand M to

different natures, physical or chemical, as well as biological. In general, selectivity

where Sj is the sensitivity of transducer j to measurand Mi. In many cases, the

Sensor Design Considerations

Selection and/or Invention of the Transduction Mechanism

Selection of Transducer Material

challenges associated with the measurement of micro- and nanomaterials properties

Physical and Chemical Considerations

A number of studies established that the melting point of a geometrically confined

depression of a particle of diameter d follows the Gibbs–Thomson equation

This relationship is only applicable if certain conditions related to the isotropy

Glass Transition Temperature

The glass transition temperature is by definition the temperature below which

Elasticity and Plasticity

displacements on the same order of magnitude as the characteristic length scales of