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 United Technologies Corporation Fire and Security Kidde Detection Technology Research Development and Engineering Colorado Springs, CO USA
Jeffrey Fortin GE Global Research Center Micro and Nano Structures Technologies Niskayuna, NY 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-achip 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 multifunctionalities on the same chip. These functionalities include, for example, onboard 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 systems 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 interface components between humans and electronic devices, or optical circuit elements 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 technology, 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, pressure sensors, accelerometers, force sensors, vibration, thickness, and chemical sensors 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 interdisciplinary integration of inorganic/semiconducting, organic/bio, and electronic/optoelectronic sensor systems. The field is constantly evolving and will continue to do so by absorbing novel materials approaches such as carbon nanotubes, high Tc superconductors, 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 February 2008
N. M. Ravindra
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 chemical 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 previously 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 departments 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 analysis, 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 determination of chemical and biological species. Compact analytical systems that have a sensor as one of the system components are becoming more important than individual 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 structures at intermediate stages of a chemical synthesis or a structural transformation. At the core of the widespread adoption of sensors are their rapidly decreasing footprint and cost and increased functionality. The miniaturization of solid-state devices in general and sensors in particular was made possible thanks to significant transformations 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 coatings, 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 significant 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 functional 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 demonstrating 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 proceedings 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 publications. He is also the co-author of a text on chemical vapor deposition polymerization of parylene and is a member of the MRS. xiii
Contents
1
Sensor Design Guidelines ........................................................................ Anis Zribi
1
2
Transduction Principles........................................................................... Jeffrey Fortin
17
3
Growth and Synthesis of Nanostructured Thin Films ......................... Yiping Zhao
31
4
Integrated Micromachining Technologies for Transducer Fabrication ..................................................................... Wei Cheng Tian
65
Applications of Functional Thin Films and Nanostructures in Gas Sensing ....................................................... Audrey Nelson
85
5
6
Chemical Sensors: New Ideas for the Mature Field ............................. Radislav A. Potyrailo
103
7
Applications of Functional Thin Films for Mechanical Sensing ......... Chang Liu
145
8
Sensing Infrared and Terahertz Regions by Functional Films ........... Magnus Willander, Victor Ryzhii, and Qingxiang Zhao
167
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 GE Global Research Center Vicktor Ryzhii University of Aizu Computer Solid State Physics Laboratory Wei-Cheng Tian GE Global Research Center 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 vocabulary 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:
[email protected] A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_1, © Springer Science + Business Media, LLC 2009
1
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 portability 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 materials, 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 performance 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 structural 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 fiberoptic 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 resonancebased signal enhancement and amplification, electro-optical devices for thermal imaging based on the pyroelectric effect, display technologies; pressure mats (piezoelectric 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 fundamentals 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-tonoise ratio photodetectors especially in the near infrared to infrared part of the spectrum 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 measurand) 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 Material
X
Transducer
Y
Y2
Y1 Amplifier
Filter
A/D
Numerical Value N
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 nanostructures are increasingly playing in high-performance transduction. Depending on the type of transducer, two basic types of sensors can be distinguished: 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 temperature 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 performance 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 technologies and a more detailed description of these parameters is provided later in this chapter.
1
Sensor Design Guidelines
5
Table 1.1 Sensor Performance Attributes Attribute Denotation Input dynamic range Response curve Sensitivity Response time Resolution Accuracy Precision Hysteresis and drift Selectivity
DM N = f(M) S t n
H Se
N Mimin
Mimax
ng
asi
re Inc
Mi
Mij
δMi
δN
sing M i
Decrea
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 sensor 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 customarily 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 = ∑ ΔMi i
∂N ∂M i
(1.1)
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 i Stransducer j =
∂N j ∂M i
(1.2)
,
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 i Stransducer
∂N p ∂M i
=
∂N p ∂N p − 2 ∂ M p −1 ∂ M p −1
...
p ∂N1 = Π S kj . ∂M i k = 1
(1.3)
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): Sfilter =
∂Y2 . ∂Y1
(1.4)
1
Sensor Design Guidelines
7
The overall sensitivity of the device to measurand Mi is defined by the following equations, ∂Y2 ∂M i
(1.5)
∂Y2 ∂Y1 ∂N . . ∂Y1 ∂N ∂Mi
(1.6)
Si total = Si total =
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 dominant 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 sensor 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): Re solution =
N N = noi Lim M i → noise Si total Si tot
(1.10)
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 transducer 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
1
Sensor Design Guidelines
9
different natures, physical or chemical, as well as biological. In general, selectivity may be quantified by p
Se = ∏ j =1 j ≠1
Stjransducer , St1ransducer
(1.11)
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 engineers 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 feasibility 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 (mechanical, 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. 2. 3. 4.
Simplicity of the transducer and robustness to failure Materials requirements dictated by operating environment Development time Cost
Selection of Transducer Material The transducer material properties are important from performance and sensor reliability 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 challenge 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 transducer 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 documented and many are being investigated (Srikar and Spearing 2003). Some of the
1
Sensor Design Guidelines
11
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 characterization to provide a sound sensor system design; in other cases initial guesstimates of the materials properties using bulk materials data and physicochemical laws combined with system-level testing, failure analyses, and design iterations will be sufficient 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, hydrogen bonds, and dipole–dipole interactions prevail at the nanoscale. In liquids confined 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): ΔTm = Tmbulk − Tm (d ) =
4s sl Tmbulk , dΔH f rs
(1.12)
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 transition 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 characteristic 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 depending 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
1
Sensor Design Guidelines
13
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 dislocations 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): ⎤ 2G 1 3 ⎡ Eˆ b = Eb + (9 L + 18 M + 8 N )⎥ ⎢K − 4a ⎣ 3 Eb ⎦ ⎤ 2G 1 ⎡1 Eˆ shear = Eshear + ⎢ ( K + 6 m ) − 1 (3 M + 4 N )⎥ , a ⎣5 3 Eb ⎦
(1.13)
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 microand 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 inhomogeneity in the z-direction can be estimated by using the PG expression shown in Equation (1.14): h( z ) = h[4p n* ( z )t * ( z )[1 + pb 0* ( z )]2 +
16 * p n ( z )b 0* ( z )] 5
(1.14)
where h=(
5 16s 2
m ) 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 proportional 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
1
Sensor Design Guidelines
15
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 dimensions 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 modifications 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 frequencies 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 agreement 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 components 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 coatings, 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 sensor 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 selective, 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 various 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 nanometre 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 particles. 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 physical 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:
[email protected] A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_2, © Springer Science + Business Media, LLC 2009
17
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 Biological Chemical Electrical Magnetic Mechanical
Optical and Radiation Thermal
Wave amplitude, phase, polarization Wave velocity Spectrum Identity, concentration, state Identity, concentration, state Current, charge potential, potential difference Field (amplitude, phase, polarization) Conductivity and permittivity Field (amplitude, phase, polarization) Flux Permeability Position, velocity, acceleration Force Stress, strain Mass, density Flow Moment, torque Stiffness, compliance Viscosity Crystallinity Wave amplitude, phase, polarization Spectrum Velocity Energy Temperature Flux Specific heat Thermal conductivity
Table 2.2 Transduction Techniques for Common Measurands Measurand Acoustic Biological Chemical Electrical Magnetic Mechanical Optical and Radiation Thermal
Most Common Transduction Techniques Utilized to Quantify the Measurand Piezoelectric Piezoresistive Capacitive Optical Piezoelectric Piezoresistive Electrical Optical Piezoelectric Piezoresistive Electrochemical Electrical Optical Electrical Optical Piezoresistive Piezoelectric Electrical – capacitive, tunneling, Optical Piezoelectric Piezoresistive Capacitive Optical Thermoelectric (Seebeck) Photosensitivity (photovoltaic, photoelectric, photoconductors, photodiodes, and phototransistors) 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, GF =
DR . R.Strain
(2.1)
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 patterned 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 piezoresistor can be located. Examples of measurands that are quantified via piezoresistance include: pressure, vibration, acceleration, and magnetic field. Once it has been decided that the measurand could be quantified via piezoresistance one must determine 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 siliconbased 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 Wire -Bond Pad
Diaphragm
Reference Cavity
Piezo- Resistor
Fig. 2.1 Side view of GE Sensing silicon pressure sensor. (Courtesy of GE Sensing.)
P1
Applied Pressure
P2
Reference Cavity 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 pressure 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.
2
Transduction Principles
21
+ + + + + + + + + + + + + + + + − − − − − − − − − − − − − − − −
(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 piezoelectric 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 piezoelectric 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 applications. 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 piezoelectric 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 moistureor 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 exposure 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 nanotubes 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 piezoresistors into the diaphragm, the diaphragm itself is used as one plate of a capacitor. 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
2
Transduction Principles
23 Pressure Ground
Si diaphragm
Reference Capacitor
Si Wafer
Fig. 2.4 Capacitive-based pressure sensor
Capacitance, pF
G-CAP2, calibrated, 10 kHz, 1 VAC, 25⬚C (77⬚F)
(a)
190 180 170 160 150 140
0
10 20 30 40 50 60 70 80 90 100 %RH
(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 piezoresistive approach is the lower power consumption of the sensor itself. Another advantage 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 measurand. 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 technique. 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 attention 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, nondestructive 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 nanosystems 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 performance of tunneling-based sensors. Liu and Kenny (2001) have demonstrated a MEMS-based high-precision, wide-bandwidth micromachined tunneling accelerometer 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.
2
Transduction Principles
25
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 thermopile. 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 junctions 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
T2
Wire B
T1
V
Fig. 2.6 Thermocouple consisting of wire of material A connected to wire of material B
26
J. Fortin V
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 measurand, or quantify an optical spectrum for chemical or biological analysis using IR, UV, or other radiation. The optical sensing techniques typically probe the material/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,
2
Transduction Principles
27
including amplitude and phase information. A radiation-sensitive detector can be used either combined with a radiative probe (e.g., Raman spectroscopy) or to measure 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 radiation primarily interacts with semiconductors via absorption processes and interference, 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 measure the intensity of a light source and is also used in power generation. In photoconduction 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 illumination incident upon the material. The resistance or conductance can then be measured 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 gratings changes as the fiber is put under stress or strain and this distance can be measured 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 measurements 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, chemotransistor, 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 underlying principle is to utilize an electrochemical reaction between the species of interest 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, polyHEMS/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 measurand 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 experimentation or innovation may be required to determine or define the best approach.
2
Transduction Principles
29
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 properties 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, TempleBoyer 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 accelerometer. 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 environments. 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 nanostructured objects such as nanoparticles, nanorods, nanotubes, nanowires, and nanoporous 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 nanostructured thin film (NSTF) is defined as an assembly of a thin layer of nanostructured 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:
[email protected]
A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_3, © Springer Science + Business Media, LLC 2009
31
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 Mechanism: physical or chemical process Environment: gas or liquid phase Temperature: high or low Other factors, such as pressure, carrier gas, plasma, substrate, etc.
Morphology: nanoparticle, nanorod, nanotubes, nanowires, nanoporous network, or mixtures Orientation: aligned or misaligned Regularity: ordered or disordered Layered structure: single layer or multilayer, single material or different materials
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 techniques 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 processes 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 fabrication 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 pressure 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 nanorod or nanowire thin films are formed. With fine-tuned growth conditions, wellaligned nanorod arrays can be fabricated. In general, the growth temperature is relatively high.
Furnace
Source materials
Carrier Gas (Ar or N2)
Growth substrate
To Pump
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 permission 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
3
Growth and Synthesis of Nanostructured Thin Films
35
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) surface 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.”
3
Growth and Synthesis of Nanostructured Thin Films
37
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 forming 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 environment. 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.
3
Growth and Synthesis of Nanostructured Thin Films
39
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 reactions, 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 section, “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 morphologies 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 increasing 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
A
B A
A A
B
B B
A A
B
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)
3
Growth and Synthesis of Nanostructured Thin Films
41
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 materials 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 semiconductor 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 sufficient 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 threeelectrode 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 deposition (Stickney 2002). This method is developed by analogy with the atomic layer deposition discussed in Section “Chemical Vapor Deposition Methods.” The surfacelimited 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 electrochemical etching of a metallic specimen, the reduction and oxidation process (redox process) takes place. All metals in contact with the solution have a pronounced 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
3
Growth and Synthesis of Nanostructured Thin Films
43 Point Charge
1 Syringe Pendant Drop
Envelope Cone y0
Secondary Bending Instability
Flat Ground Collector
(a)
−1
Grounded Plate x
0
(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 permission 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 viscosity, 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, nanoporous 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 techniques 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 deposited by thermal evaporation of Au or other metals. With the help of this electrode, 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 alternatively change the materials plated into the channel. Thus, multilayered nanorod structures can be fabricated.
3
Growth and Synthesis of Nanostructured Thin Films
45
(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 nanotubes
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 commonly 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 proceeds, 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°, obvious 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
3
Growth and Synthesis of Nanostructured Thin Films
47
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 crosssection view
Thin films deposited by oblique angle deposition have the following characteristics: 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 orientation 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 deposition; 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 simultaneously. The movement of the two motors is controlled by a computer. By changing the speed and phase of the azimuthal rotation and/or the polar rotation, the
3
Growth and Synthesis of Nanostructured Thin Films
(a)
49
(b) F
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 nanostructures 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 azimuthal rotation speed of 0.0019 rev/s and 100 total substrate revolutions. The deposition 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 columns 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 azimuthally, 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 complete 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 constant. 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 preferred 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 formation 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. 2. 3. 4. 5. 6.
Grow a layer with a thickness of L/8 without rotating the substrate. Rotate 90° at a fast azimuthal rotation rate. Repeat steps 1 and 2. Grow a layer with a thickness of L/4 without rotating the substrate. Rotate 90° at a fast azimuthal rotation rate. 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
3
Growth and Synthesis of Nanostructured Thin Films
51
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 spiral. 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 tenturn spiral film. The deposition conditions are almost exactly the same as the previous 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 programming 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
3
Growth and Synthesis of Nanostructured Thin Films
53
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 triangular 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 fabricated 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 matchsticklike 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 rotation 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.
3
Growth and Synthesis of Nanostructured Thin Films
55
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 azimuthally 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 introduce 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 incorporate the phase factor is to introduce a symmetric rotation with two different rotation 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 symmetries. 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
Growth and Synthesis of Nanostructured Thin Films
Azimuthal Rotation Angle
3
57
2π
π
φ 0
0
T/2 Program Time
T
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 revolution, 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 / 2 F|| cos ϕ F|| cos ϕ ⎤ 2 F|| ⎛ 1 1⎞ f F¢ || = ⎢ ∫ dϕ + ∫ dϕ ⎥ / t = − ⎟ sin . ⎜ wh t ⎝ wl wh ⎠ 2 ⎢⎣ −f / 2 w l ⎥⎦ f /2
(3.1)
The total time t for a complete revolution is: t=
f 2p − f + . wl wh
(3.2)
Therefore the effective flux incident angle a can be expressed by
tan a =
F¢ || F⊥
=
⎛ 1 1⎞ f 2 F sin q ⎜ − ⎟ sin 2 ⎝ wl wh ⎠ ⎛ f 2p − f ⎞ F cos q ⎜ + w h ⎟⎠ ⎝ wl
= 2 tan q sin
f 2
⎛ wh ⎞ ⎜⎝ w − 1⎟⎠ l ⎛w ⎞ 2p + ⎜ h − 1⎟ f ⎝ wl ⎠
.
(3.3)
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 substrate 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 deposition 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 geometry 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 h
q>qc
q
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
3
Growth and Synthesis of Nanostructured Thin Films
59
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 capping layer formed on top of the nanorod surface (Robbie and Brett 1997). (Reused with permission 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 fabricate this film, the sequence of the deposition is as follows. 1. 2. 3. 4. 5. 6. 7.
Open shutter. Set q = 85°, and deposit for 200 s. Change q to 75°, and deposit for 100 s. Change angle q to 20°, and deposit for 100 s. Close shutter. Change angle q from 20° to 85°. 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 incident 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 Azimuthal rotation
Motor1 Polar 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 preliminary 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 nanorod/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.
3
Growth and Synthesis of Nanostructured Thin Films
61
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 structure; 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 following 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 material can be evaporated, it can form vertical aligned nanorod structures. (4) The shape, alignment, and orientation of the nanorods can be easily changed by programming the rotation procedures. (5) The shadowing effect introduces a selfalignment 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 integrated 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 fabrication 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 polymermetal 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 selfassembled 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 nanosprings 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.
3
Growth and Synthesis of Nanostructured Thin Films
63
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 nanochannel 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 sculptured 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 mechanisms 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 evaporated-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) Onedimensional 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 transducer 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 transducer 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 structures 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:
[email protected] A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_4, © Springer Science + Business Media, LLC 2009
65
66
W.-C. Tian
In this chapter, we first introduce various micro- and nanomachining technologies 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 transferred. Other lithographic approaches in addition to optical or charged-particle-based lithographic techniques are also discussed. Different technologies for optical, chargedparticle, 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 printing, 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 maintained 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) Electron beam lithography (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), 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) X-ray (Di Fabrizio et al. 2004; Jian et al. 2003; Liu et al. 2006; Ueno 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)
Charged-particlebeam approaches
Other lithographic approaches
Mercury arc lamps and excimer lasers are conventional light sources for lithographic 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 lithography 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 diffractionlimited 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 lithography 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 lithography (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 electrons) are used in conjunction with a charged-particle sensitive resist. These techniques 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 scattering 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 immersion 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.
4
Integrated Micromachining Technologies for Transducer Fabrication
69
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 permission 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 addition, 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 optical/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 ablation, 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 technologies. 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
4
Integrated Micromachining Technologies for Transducer Fabrication
71
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 microstructures fabricated by chlorine-based etching are very vertical and no sidewall passivation 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), tetramethylammonium 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 structures 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 building 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 blasting (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 physically 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 typically 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 micrometer 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 materials 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 developed 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 nanostructures over large areas. Step and flash imprint lithography (SFIL) using photopolymerization was developed 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 transparent 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 patterns now become the etch mask and the patterns can be transferred to the underlying 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.
4
Integrated Micromachining Technologies for Transducer Fabrication
73
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), polycarbonate (PC), polyethylene (PE), polyetherether-ketone (PEEK), PMMA, polystyrene (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 microfluidics (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 polymer 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 section, 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, semiconductor-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 deposited 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
Semiconductor-based
MEMS-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) 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) 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)
4
Integrated Micromachining Technologies for Transducer Fabrication
75
Other substrates, such as alumina for conductance measurement and lithium niobate (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 electrodes 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 typically 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 measured 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 parameters (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 silicon 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 discussed 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 nanostructures with interdigitated microelectrode (IME) transducers. The IME transducer is fabricated with metallization followed by lithography. This IME substrate is subsequently immersed into the prepared coating sol or the solution of mixed nanoparticles 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 nitroaromatic compounds using nanoparticles/thiols on IME transducers have been demonstrated (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 dendrimer/metallic nanoparticles film for biosensing (Goulet et al. 2005). A combination 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).
4
Integrated Micromachining Technologies for Transducer Fabrication
77
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) Immersion (Han et al. 2005) Layer-by-layer (LBL) assembly LBL (Hammond 2004) LBL (Kim et al. 2004) LBL (Lutkenhaus et al. 2005) LBL (Goulet et al. 2005) LBL and photolithography (Cui et al. 2004) Other self-assembly Magnetically aligned (Hernandez et al. 2005) Biological-based self-assembly (Zhao G et al. 2005) Nanostructured deposition Electrochemical Deposition (Park et al. 2005) Deposition metal–insulator–metal ensemble (Smardzewski 2004) Sputtering (Ippolito 2005) Sputtering (Zhao Z et al. 2005) Sputtering (Mardare et al. 2005) Plasma enhanced chemical vapor deposition (PECVD) for carbon-silicon and sputtering for metal (Takeno et al. 2005) Pulsed laser deposition (Modi and Lacy 2005) Pulsed laser deposition (Vainos 2004) Glancing angle deposition (Kim DS et al. 2006) Thermal evaporation of metallic tungsten oxide powder followed by thermal annealing (Jayatissa and Gupta 2005) MBE and lithography (Lapicki et al. 2005) Polymerization Electropolymerization (Bangar et al. 2005; Fan and Lu 2005; Hammond 2004) Spin coating and UV polymerization (Schmidt and Haupt 2005) Spotting Carbon nanotubes (CNTs) were dissolved in ethanol and sprayed by air brush (Sayago et al. 2005) Ink jet spotting (Braun et al. 2006) Pipetting (Zribi et al. 2005)
Titanium oxide (TiO2) Nanoparticles and thiols Various materials Polyelectrolytes and TiO2 nanoparticles Polyacrylic acid (PAA); polyethylene oxide (PEO); dye Dendrimer/metallic nanoparticles on glass Polystyrene nanoparticles Iron palladium (FePd) alloy nanowires Zirconium dioxide (ZrO2) nanoparticles with heme proteins on functional glassy carbon electrode Nanoporous platinum oxide (PtO) Alkanethiol–Au nano cluster Zinc oxide/indium oxide (ZnO/InOx) on XZ lithium niobate (LiNbO3) substrate 60% Palladium (Pd) 40% gold (Au) Chrome (Cr)-doped titanium oxide Metal-containing diamondlike nanocomposite ZnO nanorods Purely epitaxial nanocomposite Silicon dioxide Tungsten oxide (WO3) Indium antimonide (InSb) Monomer nanowires Molecularly imprinted polymer
Single-walled carbon nanotube (SWCNTs) with palladium doping Proteoliposomes 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 magnetically 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 peroxide sensing (Zhao G et al. 2005).
Nanostructure Deposition Nanostructures can be formed using various processes, such as electrochemical 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 lithography, etching, and metal contact deposition is used to fabricate this sensor platform.
Polymerization Polypyrrole films doped with poly-aminobenzene sulfonic acid (PABS) functionalized 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.
4
Integrated Micromachining Technologies for Transducer Fabrication
79
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 addressability. By simply adding biomolecules in the solution, biofictionalization of conducting 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 detection with quantitative data have been reported by Braun et al. (2006). Application of proteoliposomes onto MEMS cantilever arrays using ink jet spotting was demonstrated. The authors claimed that this technique can be potentially used to measure 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 gassensing 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 nanotubes 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 applications. 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-generation 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.
4
Integrated Micromachining Technologies for Transducer Fabrication
81
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 electroforming. 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 nanoscale. 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 channel 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 nonchemically 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 microfabrication. 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 assemblies. 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.
4
Integrated Micromachining Technologies for Transducer Fabrication
83
Park S et al. (2005) pH-sensitive solid-state electrode based on electrodeposited nanoporous platinum. 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 Micromechanics 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 femtosecond 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 lithography. 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 microstructures. 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 transistors. 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 lithography 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 instrumentation used for gas sensing, such as optical nondispersive infrared (NDIR) sensors. Because of recent developments in thin-film and nanostructure technologies, 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 bringing 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:
[email protected]
A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_5, © Springer Science + Business Media, LLC 2009
85
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 developed 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 wavelength components with a grating. These dispersive spectrometers have a fundamental 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 pioneered 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 nondispersive 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 improvement factor known as the multiplex or Fellgett’s advantage. In addition the interferometer 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 nondispersive spectrometers are widely used in laboratories for the qualitative and quantitative 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 wavelengths 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 wavelength 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 electromagnetic 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 quantum 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 excitations 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 absorbing the light if the wavelength of the light matches a wavelength in the vibration– rotation band.
5
Applications of Functional Thin Films and Nanostructures
89
Typical incandenscent lamp radiation intensity spectrum
Arbitrary units
1.2E-15 1E-15 8E-10
Mid IR
6E-16 4E-16 2E-16 0
0
1
2
3
4
5
6
7
8
9
10
Wavelength in microns
Transmittance%
Visible 100 90 80 70 60 50 40 4.2
H2O
CO CO2 4.4
N2O 4.6
4.8
5
5.2
5.4
5.6
Wavelength 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 particular absorption column. When L = 0 or L = ∞, transmission has no sensitivity to
90
A. Nelson 0
T(L) dT/dC(L)
1
−0.000005
Transmission
−0.00001 0.8
−0.000015
0.6
−0.00002 −0.000025
0.4
−0.00003 0.2
−0.000035
0 0
25
50
75 100 125 Path length in mm
150
175
Sensitivity dT/dC measured in 1/ppm
1.2
−0.00004 200
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 giving 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 (C ) =
∫ i (l )e 0
( − a ( l )CL )
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 simultaneously with I(C) in one gas volume.
5
Applications of Functional Thin Films and Nanostructures
91
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 technology 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 technology 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 operating 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
Sample Cell
Reference Cell Detector
Fig. 5.4 A NDIR system
IR Source
Mirror
Chopper Motor
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 shortcomings: 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 thermal 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 temperature 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 junction is kept at a known temperature by bonding it to a mass with a stable temperature.
5
Applications of Functional Thin Films and Nanostructures
93
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 polarization 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,
(Wolfe WL, Zissis GJ (1985)) where < DF 2 > = 4 kTd 2 GDf
(5.6)
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 temperature 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
Sources
IR lamp
Thermal source
Life-time Modula(h) tion
40,000
Nanoamorphous 25,000 carbon Films
λ Range (µm)
<2 Hz
0.2–5
<100Hz
1 – 20
Photo
Vendor
International Light Technologies [ILT] Intex
[Intex]
Surface textured thin metal filament
25,000
<10 Hz
2 – 20
ICx Photonics [ICx Photonics]
Lead-salt Diode Lasers
10,0001
10kHz2
3 – 30
Laser Components [Laser Components]
Quantum Antimonide Laser Lasers
Quantum Cascade Lasers
2,0003
10KHz 2–4
7,000
10KHz 4 – 12
Roithner LaserTechniK GmbH [Roithner Laser] Alpes Lasers [Alpes Lasers]
1
The Laser Guidebook by Jeff Hecht, Chapter 19, p. 280. 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). 2
detectors present a great challenge, because of the limitations of the narrow bandgap 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 important characteristics.
5
Applications of Functional Thin Films and Nanostructures
95
Table 5.2 Various Infrared Detectors Spectral Operating Detectivity Da Response (mm) Temperature (K) (CmHz1/2/W) Company
Types
300
1.8 × 108
300
2 × 108
8–14
300
1 × 109
2–6
213
2 × 109
1.5–5.8 2–16
300 77
1 × 108 2 × 1010
ThermoThermopiles 2–15 detectors Pyroelectric 2–20 Bolometer Quantum InSb detectors PbSe HgCdTe
Perkin-Elmer (Perkin) IGM&I Co. Ltd (IGM&I) Raytheon (Raytheon) Hamamatsu (Hamamatsu) Hamamatsu 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 bandpass 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 filters
Transmitting only a certain wavelength range
Notch Filters
Eliminating only a certain wavelength range
Low Pass Filters
Transmitting only wavelengths below a certain value
High Pass Filters
Transmitting only wavelengths 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-
5
Applications of Functional Thin Films and Nanostructures
97
Table 5.4 Common Infrared Coating Materials (Harrick)
Percent transmission
Sapphire 0.1-5.0
100
Percent transmission
1.76
100
Percent transmission
Refractive index@ (mm)
100
Percent transmission
Wavelength Range (mm)
100
Percent transmission
Material
100
80 60 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
BaF2
0.2-11
80 60 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
3.42
Silicon
1.1-10
80 60 40 20 0 1 .2 .3 .4 .5 .6 .7 .8 .9 1.0
2.0 3.0 4.0 5.0
10
20
Wavelength in microns
2.49
ZnSe
0.6 – 20
80 60 40 20 0 1 .2 .3 .4 .5 .6 .7 .8 .9 1.0
2.0 3.0 4.0 5.0
10
20
Wavelength in microns
1.54
KBr
0.5 - 25
80 60 40 20 0 1 .2 .3 .4 .5 .6 .7 .8 .9 1.0
2.0 3.0 4.0 5.0
10
20
Wavelength in microns
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 bandpass 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.
Diffusion Membranes
Incandescent Infrared Source
Custom Designed Infrared Filters Reference Target Gas
Dual Beam Micro-Machined Thermopile Detector
1.25" Patented Waveguide
Microprocessor
Fig. 5.5 Telaire NDIR system
5
Applications of Functional Thin Films and Nanostructures
99
• 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 background 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 factor 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 coefficient 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.
5
Applications of Functional Thin Films and Nanostructures
101
1.8E-05 Ambient N2 Purge
Intensity (a.u.)
1.4E-05
1.0E-05
6.0E-06
2.0E-06 3
3.5
4 4.5 Wavelength (um)
5
5.5
Fig. 5.6 RCLED spectrum
Table 5.5 Different System Configurations Source
Detector
Modulation
Peak Current
Avg. Power Consumption
IR lamp LED LED
Thermopiles Pyroelectric Microphone
1–5 Hz 0.25–1 kHz 1–100 kHz
100 mA 100 mA 100 mA
20–80 mW 50–120 µW 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-feedback 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 signaltransduction 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 everexpanding 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: [email protected] A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_6, © Springer Science + Business Media, LLC 2009
103
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 material is applied onto a suitable physical transducer to convert a change in a property 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., spectroscopic, 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 miniaturization (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.
Transducer Sample introduction
Sensing material
Radiant energy Electrical energy Mechanical energy Thermal energy
Data processing
Species identity Species concentration
A B C D Sample components
Fig. 6.1 Main components of a modern chemical sensor system
Nanoparticle metal oxide
Polymeric film
Polymeric film
Polymeric film formulated with organic fluorescent dye Metallo-porphyrin dye on solid supports Formulated catalytic composition Formulated catalytic composition
Electrical
Mechanical
Mechanical
Radiant
Thermal
Thermal
Radiant
Composite conducting polymer
Type of Sensing Material
Electrical
Type of Energy Involved in Transduction Principle
CO, NOx, O2, volatile organic compounds
Homeland security and defense applications
Analyte/Applications
Colorimet-ric sensor film array Catalytic combustion sensor Catalytic combustion sensor
Combustible higher hydrocarbons
Combustible hydrocarbons
Volatile organic compounds
Array of surface acoustic Nerve and blister agents wave devices Micro-machined strain Water vapor gauge Container-integrated pH, dissolved O2, pCO2 disposable patch
Micro-hotplate
Array of inter-digitated electrodes
Transducer Design
Table 6.1 Representative Examples of Commercially Available Chemical Sensor Systems
ChemSensing, Inc., Northbrook, IL, USA, www.chemsensing.com Delphian Corp., Northvale, NJ, USA www.delphian.com Nemoto & Co, Milan, Italy www.nemototech.com
Smiths Detection-Pasadena, Inc. (former Cyrano Sciences, Inc.), Pasadena, CA, USA www.smithsdetection.com Microchemical Systems SA, Corcelles, Switzerland www.microchemical.com MSA, Pittsburgh, PA, USA www.msanorthamerica.com Hygrometrix Inc., Alpine, CA, USA www.hygrometrix.net Fluorometrix Corp., Stow, MA, USA www.fluorometrix.com
Company
6 Chemical Sensors: New Ideas for the Mature Field 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 nanofabrication 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 highthroughput 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 methodologies (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 resolution 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 functionalities 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 challenges 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 Requirements Dynamic range Ergonomic design False positive rate Initial cost Long-term stability Maintenance simplicity Multicomponent detection Multiple operation modes Operation cost Operation simplicity Power consumption Probability of detection Response speed Response reversibility Robustness Selectivity Self-calibration Sensitivity Shelf-life Size Total (out of 20)
Sample Introduction
Sensing Material
Transduction Principle
Data Analysis
X X X X X X X
X
X
X
X X X X X X
X X X X X X X
X X X
X X X X X X X X X X X 18
X X X
X X X X X X X 14
X X
X X
X X 15
X X X X X X
13
Data analysis
Transducer
– Preconcentration for detection limit improvement
– Reliable quantification of – Analysis of dynamic signatures for four or less analytes in their selectivity and stability mixtures with sensor array improvement of partially selective sensing – Analysis of multivariate signatures for selectivity, signal-to-noise, films and a single transducand stability improvement tion principle
–
– –
Sensing material – –
– Particulate contamination – Lack of representative sample for trace analysis
Sample introduction
Sensor-Design Aspects
– Time-domain modulation for selectivity improvement – Integration into periodic self-cleaning sensor system Film poisoning – Temperature-stabilized operation (T > ambient) Water condensation – Temperature, gas-flow modulation to facilitate more reversible and selective response Corrosion – Transducer design for higher-order sensor response through hyphenated Signal-to-noise decrease with transduction techniques and time-, reduced transducer size temperature-, sampleSophisticated readout with flow-modulation methods for selecreduced transducer size tivity, signal-to-noise, stability improvement
Challenges
Sub-systems
Gas-Phase Chemical Sensing
– Covalent reagent attachment to matrix – Cartridges of disposable sensors
– Reagent leaching – Film poisoning
– Transducer design for higherorder sensor response through hyphenated transduction techniques and time-, temperature-, sample-flow-modulation methods for selectivity, signal-to-noise, stability improvement – Reliable quantification – Analysis of dynamic signatures of four or less analytes for selectivity and stability in their mixtures with improvement sensor array of partially – Analysis of mltivariate signatures for selectivity, signal-toselective sensing films noise, and stability and a single transducimprovement tion principle
– Corrosion – Signal-to-noise decrease with reduced transducer size – Sophisticated readout with reduced transducer size
– Film delamination
– Protective coatings – Replaceable sampling components
Sensor-Design Aspects
Liquid-Phase Chemical Sensing – Corrosion – Scale deposition – Lack of representative sample for trace analysis
Challenges
Table 6.3 Examples of Challenges and Sensor-Design Aspects Associated with Gas- and Liquid-Phase Chemical Sensing
108 R.A. Potyrailo
6
Chemical Sensors: New Ideas for the Mature Field
109
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 automotive 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 advantage 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 automatic 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 sensitive 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 information about the sensor and the object onto which the sensor is attached. Highly efficient transducers of mechanical energy are also available as massproduced 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 instrumentation, and, allows a compact design for device applications.
110
R.A. Potyrailo 1
(a)
2
prongs
(b)
(c)
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 piezoresistive 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, efficient 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
6
Chemical Sensors: New Ideas for the Mature Field
111
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 applications 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, mechanical, 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 electrodes makes sensors based on electrical energy transduction among the most employed tools in chemical sensing. Typical electrode designs incorporate the electrical 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 identification 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.
6
Chemical Sensors: New Ideas for the Mature Field
113
(a)
Sensing film RFID sensor
RF
Antenna Sensing film Inert substrate
CF
(b) Antenna RA LA CA
RFID Reader
Microchip
RFID tag
(c) Fp Zp F2
Zre 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 transduction 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 capacitance 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 materials. Examples of formulation components in Nafion films include conducting polymers (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 conductivity can also be employed for multianalyte detection with a single RFID sensor.
6
Chemical Sensors: New Ideas for the Mature Field
115
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 complex impedance were calculated (Fig. 6.3c). These parameters included maximum 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 imaginary part of the complex impedance (F2). Because F1 and F2 are related to different 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 concentrations 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 relation 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 impedance 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 determinations 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
(a)
10 0
Delta Zp (Ohm)
−10 −20 −30 H2O
−40
MeOH −50
EtOH
−60
ACN
−70
0
1
0.5
2
1.5
3
2.5
3.5
Time (h) 1.5
(b) 1
MeOH 0.2
ACN 0.2
EtOH 0.15 0.2 0.15 0.1 0.1 0.15 0.07 0.07 0.1 0.07 0.04 0 0.04 0.1 0.04 0.02 0.02 0.07 0.02 −0.5 0.04 0.02 0 −1
PC 2
0.5
0.15
H2O −1.5 −2 −2
0.2
−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
6
Chemical Sensors: New Ideas for the Mature Field
117
the smallest measured vapor concentration (0.02 P/P0) to be 600 ppb. Upon a careful 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 positionindependent (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 illustrated 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 measurements, 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 interested 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 position 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
Distance (mm)
20
Z Direction
X Direction
15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Positions
RFID Sensor Signal (Arb. Units)
(c)
45%RH
45%RH
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 positions 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
(b)
(a)
20000 F1 Shift (Hz)
100000 50000 0
−50000 0
Frequency Shift (Hz)
Frequency Shift (Hz)
150000
F2 Shift (Hz)
0 −20000 −40000 −60000 −80000 −100000 −120000
80 160 240 320 400 480
0
80 160 240 320 400 480
Time (min)
Time (min)
(c)
(d) Fp Shift (Hz)
0
−20000 −40000 −60000
Zp
150
20000
Zp
Frequency Shift (Hz)
40000
100
50 0
80 160 240 320 400 480
0
80 160 240 320 400 480
Time (min)
Time (min)
(e) 5 4
Principal Component #2
3 2 1 0 −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 frequency 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)
4
99
8 12 100
97
Zp (Ohm)
Zp (Ohm)
98
96 95
28
96 94 92
94
20
98
0
0
44
10 20 30 40 50 NaCl conc. (ppb)
200
400 600 Time (s)
800
1000
(b) 1.2 bare sensing film
Zp normalized
1 0.8 0.6 0.4 0.2 0
NaSO4
(c)
HCl NaCl Ionic Species
KH2PO4
60000
Freq. shift (Hz)
50000 40000 30000 20000 10000 0 0
2 4 6 8 Acetone conc. %vol.
10
Fig. 6.7 Applicability of RFID sensors for monitoring in liquids. (a) Analysis of ion concentrations 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 polysulfonecoated 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,
6
Chemical Sensors: New Ideas for the Mature Field
121
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 sensors 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 transducer 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 interactions (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 demonstrated 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
N
N
N
O
O
O O
Scheme 6.2
O
x
O O
N O
O
Si-O Si O -Si ~10
y
122
R.A. Potyrailo
Sensor response (− Hz)
40
(a) 5.3
30 2.4
20
1.5 10 0.5 0 200 400 600 800 1000 Vapor concentration (ppb)
0
5.5
(b)
y = 4.818−0.493x R2 = 0.999
5
log (K)
4.5 4 3.5 3
conventional polymers
2.5 −2
−1
0
1
3
2
Log (Vapor concentration, ppm)
Signal Intensity (counts)
(c)
C
O
Three-years old
N
Si
Fresh
1000
800 600 400 200 Binding Energy (eV)
0
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-billion 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
6
Chemical Sensors: New Ideas for the Mature Field
123
presented in Fig. 6.8a. We observed that the sorption isotherm of new sensor polymers to different chlorinated and other types of vapors was nonlinear. The origin of this nonlinearity is likely to be from a combination of different vaporsorption mechanisms with contributions from both Langmuir- and non-Langmuirtype 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 concentrations was negligible, <1 Hz because partition coefficients of different conventional polymer materials for TCE and other chlorinated volatile organic vapors of environmental 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 100fold enhancement of the partition coefficient of our new material for detection at ppb levels when compared to conventional currently used sensor materials such as phenylmethyl-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 material. 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 thickness 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 selectively 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 piezoresistive readout. Often, cantilever response is measured in air on an atomic force microscope (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 readout 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 converted 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 6
1
4
Scores on PC# 2
2 0
3
-2
4
-4
5 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
6
Chemical Sensors: New Ideas for the Mature Field
125
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 properties 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), aminopropylsiloxane–octadecylsiloxane copolymer (Oprea et al. 1999), polyethyleneimine (Korsah et al. 1998), fluorinated polyimide (Hoyt et al. 1998), and tetrakis(hydroxyethyl)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 increasing 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 operating 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
Sensor Response (mV)
6400
(a)
5600 4800 4000 3200 2400 1600
CO2 increase CO2 decrease
800 0
0
20 40 60 80 CO2 Concentration (%)
100
Sensor Response (mV)
4000
(b)
3000 2000 1000 0
−1000 −2000 −3000
0
2000
4000 6000 Time (s)
8000
Sensor Response (mV)
250
(c) 200 150 50
100
25
50 0
0
0
0
4000
10000 20000 30000 CO2 Concentration (ppm)
8000
40000
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.
6
Chemical Sensors: New Ideas for the Mature Field
127
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 prescribed 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 transducer, 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 commercial 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 engineered 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 nanocrystals incorporated into a polymer matrix to overcome a photobleaching limitation of traditional organic reagent-based sensor materials. We have found that when different 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 nanocrystals 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 surrounding the nanocrystals, their size, and surface oxidation state. The sensitivity of these sensors was defined as a signal change upon exposure to the known concentrations 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 electronic 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 nanocrystals 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 nanocrystals 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-
6
Chemical Sensors: New Ideas for the Mature Field
129
(a) 1500
PL (counts)
5.6 nm 1000
500
2
1
2
1
2
1
2.8 nm 0
(b)
0
50
100 150 Time (min)
200
5
PC 3 ( 5.09%)
0 −5
1
2
−10 −15 −20 −15
(c) 1400
−10
−5
0 5 10 PC 1 (53.96%)
15
20
22 230
1000 11 PL (counts)
PL (counts)
1200
800 600 400
220
190 180
0
2
4
22
200
200 0
11
210
6
0 0.5 1 1.5 2 2.5 3 3.5 Time (h)
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 continuous 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 generating highly stable multivariate response patterns from different-size CdSe nanocrystals incorporated in PMMA. Our studies are in progress to better understand the effects of polymer matrices. Such work promises to complement existing solvatochromic 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 detection, 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 Reflective layer
Optical disk 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 signal acquisition and drive control
Analog signal
(b)
Disk Label Reflective layer Sensor film Detected laser light
(c)
Optical phenomena Amplitude
Polarization
Phase
in de x ct fra Re
Th
ick
ive
ne ss
iv ity R
ef le ct
g rin te at Sc
Ab
so
rb an ce
Parameters of sensor film applicable for chemical / biological quantification in an optical disk drive
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 optical 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 controlled 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 quantitation 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 wavelength 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
6
Chemical Sensors: New Ideas for the Mature Field
133
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 measurements 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 chromogenic dyes immobilized in polymeric films (Seitz 1988; Wolfbeis 1991). The calculated 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+ determinations 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¢-diethyl4,4¢-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 introduced 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
R.A. Potyrailo
Normalized DVD Signal
134
(a)
100 80 60 40 20 0
10 100 Calcium Conc. (ppm)
1000
200 400 600 800 Chlorine Conc. (ppb)
1000
Normalized DVD Signal
1
(b)
100 75 50 25 0
Normalized DVD Signal
0
76
(c)
72 68 0 10 20 30 40 50 60 70 % RH
64 60 1
56
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 determinations 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.
6
Chemical Sensors: New Ideas for the Mature Field
135
In conclusion, Lab-on-DVD technology coupled with diverse indicator film chemistries 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 nanoassembled sensing materials promise to provide previously unavailable sensing capabilities. 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 sensing applications. However, the compatibility of established integrated-circuit technologies (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 miniaturization 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.
6
Chemical Sensors: New Ideas for the Mature Field
137
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, editors. 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 hydrogenbonded 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 solvent 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 nanoparticles. 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 monitor 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 sensing 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 tetramethylammonium 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 recognition. 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 contributions to thickness shear mode and surface acoustic wave vapour sensor responses using partition coefficients. Faraday Discuss. 107:259–283. Grate JW, Kaganove SN, Bhethanabotla VR (1998) Comparisons of polymer/gas partition coefficients 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 sensing based on digital color analysis. Anal. Chem. 72:465–474. Hirschfeld T (1985) Instrumentation in the next decade. Science 230:286–291.
6
Chemical Sensors: New Ideas for the Mature Field
139
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-temperature measurement of CO2 and relative humidity. Anal. Chem. 70:2137–2145. Hruschka WR, Massie DR, Anderson JD (1983) Computerized analysis of two-dimensional electrophoretograms. 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 ionconducting 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 sensors: 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 permselective 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 transducers. Sens. Actuators B 120:353–361. McCurley MF, Seitz WR (1991) Fiber-optic sensor for salt concentration based on polymer swelling coupled to optical displacement. Anal. Chim. Acta 249:373–380. McGrath JE, Dunson DL, Mecham SJ, Hedrick JL (1999) Synthesis and characterization of segmented 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 monodisperse 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 sensors. 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.
6
Chemical Sensors: New Ideas for the Mature Field
141
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 combinatorial 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 structurally 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 highthroughput 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 proteins 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 immobilized 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 humidity-sensing polymers. Sens. Actuators B 40:211–216. Wang J (2002) Electrochemical detection for microscale analytical systems: A review. Talanta 56:223–231.
6
Chemical Sensors: New Ideas for the Mature Field
143
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. Proceedings 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 motionsensitive 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: [email protected] A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_7, © Springer Science + Business Media, LLC 2009
145
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 experience 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 measuring 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 profile 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 functional 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 electrostatic 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 advantage 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, functional thermal electric-based temperature sensors may be used for measuring temperature 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 exhibits 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 displacement 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 characteristics 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 conductive 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
7
Applications of Functional Thin Films for Mechanical Sensing
149
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 materials 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 deposition. Using this method, a relatively large thickness (e.g., 7 mm) can be reached
Table 7.1 Properties of selected piezoelectric materials
Material ZnO PZT-4 (PbZrTiO3) PZT-5A (PbZrTiO3) Quartz (SiO2) Lithium tantalate (LiTaO3) Lithium niobate (LiNbO3) PVDF **Data not available
Relative Permitivity Young’s Density (Dielectric Constant) Modulus (GPa) (kg/m3)
Curie Coupling Temperature Factor (k) (°C)
8.5 1,300–1,475
210 48–135
5600 7500
0.075 0.6
** 365
1,730
48–135
7750
0.66
365
4.52
107
2650
0.09
**
41
233
7640
0.51
350
44
245
4640
**
**
13
3
1880
0.2
80
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 carbonate and bismuth oxide as the bonding agent. After screen printing, the deposited materials are dried and then fired at high temperature for densification. 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 sputtering, 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 semiconductor 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
7
Applications of Functional Thin Films for Mechanical Sensing contact region with metal leads (heavily doped)
151
resistor (moderately doped)
A
A’ desired resistor pattern
doping contact region
mask shield
dopant atoms
(a)
mask shield doping resistor
mask shield
A-A’ mask shield
contact doping
mask shield
heavy doping resistor doping
mask shield
(b) A-A’ metal leads
ohmic contact mask shield
metal
moderate doping ohmic contact
(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
Log resistivity (Ω cm)
10 8 6 4 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 concentration. 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 material. 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 sufficiently covered in textbooks such as Gere and Timoshenko (1997). In this chapter we simply provide the most commonly used formula for analyzing stress and
7
Applications of Functional Thin Films for Mechanical Sensing
153
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 twodimensional 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 summarized 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 diving 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 rotational 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 displacement is to solve a second-order differential equation of a beam: EI
d2 y = M ( x ), dx 2
(7.1)
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 conditions 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 displacement (arc angle).
7
Applications of Functional Thin Films for Mechanical Sensing
155
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 Max value of θ =
Fl 2 2EI
Maximum vertical displacement occurs at the free end Fixed-free beam under point loading at the free end.
Max. value d=
Fl 3 3EI
q at the free end equals zero due to guided boundary condition. Maximum vertical displacement occurs at the free end
Fixed-guided beam under point loading at the free end.
Max. value d=
Fl 3 12EI
Maximum vertical displacement occurs in the middle Max. value d=
Wl 3 192EI
A fixed-fixed beam with point loading applied at the center. Maximum q occurs at the end of the beam Max q =
Ml EI
Maximum vertical displacement occurs at the free end Fixed-free beam under a torque loading. The position of the torque along the length is not relevant.
Max d=
Ml 2 2EI
Beam Stress Analysis The analysis of intrinsic stress in a beam under flexural loading can be very complicated 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 location (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 distribution 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
7
Applications of Functional Thin Films for Mechanical Sensing
157
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): e max =
M ( x )t FLt = . 2 EI 2 EI
(7.2)
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 + = , ∂x 4 ∂x 2 ∂y 2 ∂y 4 D
(7.3)
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 D=
Et 3 . 12(1 − n 2 )
(7.4)
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 wcenter
b σcenter a a/b β1 β1 α
1.0 1.2 1.4 1.6 1.8 2.0 0.3078 0.3078 0.3078 0.3078 0.3078 0.3078 0.3078 0.1386 0.1386 0.1386 0.1386 0.1386 0.1386 0.1386 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:
wcenter =
apb 4 , Et 3
(7.5)
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:
s mas =
b1 pb 2 , t2
s center =
b 2 pb 2 , t2
with the values of b1 and b2 listed in the table as well.
(7.6) (7.7)
7
Applications of Functional Thin Films for Mechanical Sensing
159
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 biological 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 membrane 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 responsible for hearing (human cochlea), flow sensing (insects, spiders, and fish), vibration 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 highperformance 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 development 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 characteristics. 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, potentially reducing the costs of final sensor products. Recent development of an artificial haircell sensor is discussed below. The schematic 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:
e=
6Fl Flt = , 2EI Ewt 2
(7.8)
7
Applications of Functional Thin Films for Mechanical Sensing
161
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 resultant 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 patterned 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.
7
Applications of Functional Thin Films for Mechanical Sensing
163
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 packaged 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
Sensor Output (V)
1 0.8 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
7
Applications of Functional Thin Films for Mechanical Sensing
165
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 systems 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 prepared 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 electromagnetic (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 elementary 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: [email protected] V. Ryzhii Computer Solid State Physics Laboratory, University of Aizu, Aizu-Wakamatsu, 965-8580, Japan
A. Zribi and J. Fortin (eds.), Functional Thin Films and Nanostructures for Sensors, Integrated Analytical Systems, DOI: 10.1007/978-0-387-68609-7_8, © Springer Science + Business Media, LLC 2009
167
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 twodimensional 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 quantum 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 lowdimensional semiconductor structures, in particular, in quantum well (QW) structures. 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 conduction 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 electrons 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 photodetectors 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,
170
M. Willander et al. Grating
(a)
(b)
(c)
Emitter
QWs QWIP
Collector
Emitter
QRs QRIP
Collector
Emitter
QDs
Collector
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-nN-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 applications 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 current across the active region depends on the applied voltage and the injection conditions (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
8
Sensing Infrared and Terahertz Regions by Functional Films
(a)
171
(b)
Emitter Emitter QDs (QRs)
QDs Collector
(QRs) 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 barrier at the emitter edge of the active region (in photodetectors with tunneling injection) 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 components: 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 injection. 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, particularly in the lateral direction (virtually uniform in QWIPs and strongly nonuniform 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
172
M. Willander et al.
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 contrast, 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 different. 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 temperature 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 determined by the activation energy ea = ei − eF. Here ei and eF are, respectively, the ionization 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⎥ ≈ , mk BT m ⎢ ⎥ ⎣ ⎦
(8.1)
where Σ is the average electron sheet concentration in QWs (or, in the following, 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 .
8
Sensing Infrared and Terahertz Regions by Functional Films
εa
εi
173
(a)
(b) εa
εi
QD shells
QR subband bottoms
εa
εi
εa
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 spectrum, the Fermi energy can be estimated using the following formula,
e
(QW ) F
Σ
Σ
⎡ ⎤ p 2 p 2 ⎢ ⎥ ) − 1⎥ ≈ = k BT In ⎢exp( 2 mk BT a 2 ∑ QR ⎢ ⎥ m a ∑ QR ⎣ ⎦
(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 narrow 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:
e F(QR ) ≈
p 2
Σ
8m ∑ QR
2
.
(8.3)
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.
174
M. Willander et al.
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 2 example, Helm et al. (1994). This means that the product a ∑ QD should not be too close to unity. As a result, the following expression is obtained.
e
⎡ ⎤ p 2 ∑ p 2 ∑ ⎢ ⎥ ) −1 ≈ . = k BTIn exp( ⎢ ⎥ ma 2 ∑ mk BTa 2 ∑ QD QD ⎣ ⎦
(QD ) F
(8.4)
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 thermal dark current,
j th ≈
eGth , p
(8.5)
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. 2 Considering Equations (8.1), (8.2), (8.4), and (8.5) and introducing ∑ T = mk BT / p , the density of thermal dark current can be expressed byt
j th(QWIP ) α
≈
j
≈
(L − QRIP ) th
p
(QW )
p
(QW )
exp(
∑
∑
)exp(−
T
a 2 ∑ QR ⎡ ⎢exp( α (QR ) ⎢ p ∑T ⎢⎣
a 2 ∑ QR p
1
⎡ ⎤ ⎢exp( ∑ ) − 1⎥ exp(− ei ) ⎢ ⎥ k BT ∑T ⎣ ⎦
1
(QR )
exp(
∑
∑
T
a 2 ∑ QR
(8.6)
ei ) k BT
∑ a 2 ∑ QR
⎤ e ) − 1⎥ exp(− i ) ⎥ k BT ⎥⎦
e )exp(− i ) k BT
(8.7)
8
Sensing Infrared and Terahertz Regions by Functional Films
j
(L − QDIP ) th
≈
⎤ a 2 ∑ QD ⎡ ∑ ⎢exp( ⎥ exp(− ei ) ) 1 α − (QD ) 2 ⎢ ⎥ k BT p ∑ T a ∑ QD ⎣ ⎦
a 2 ∑ QR
exp(
(QR )
p
175
∑
∑ T
a
2
∑
)exp(− QD
(8.8)
ei ). k BT
If the densities of QR and QD arrays approach their maxima ∑ QR = a and ∑ 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 , and Equations = p = p −2
(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
( S − QRIP ) th
j
α
a 2 ∑ QR p
( QR )
p exp(
∑
2
8∑ T ∑ QR
exp( −
ei ). k BT
(8.9)
In the case of relatively small QDs having a single quantum shell (with maximum two electrons), considering that e a = e i (see Fig. 8.3), for the thermoemission rate one obtains:
jth( S −QDIP ) α
∑ 2 p
( QD )
∑
exp( − QD
ei ) . k BT
(8.10)
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 probability. Because the energy distribution of mobile electrons is determined by the electric field, which can give rise to a significant electron heating (Ryzhii and
176
M. Willander et al.
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 quantumlevel 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 conditions 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),
p
( QD )
⎡max N 2 ∑ QD − ∑ ⎤⎦ e ∑ ∝ (a 2 ∑ QD ) ⎣ exp( − ) max N ∑ QD e CQD ∑ QD
(8.11)
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 probability 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 ( QD ) tends to zero when ∑ approaches the maximum value allowed by the p
Pauli principle, that is, to max N ∑ QD . Equation (8.11) also shows that p as a function of ∑ contains an exponential factor associated with the effect of the ( QD )
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 decreasing p
( QD )
–
∑ and they are steeper than this current in QWIPs. Strongly ( QR ) ∑ and p – ∑ dependences result in an even more
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.
8
Sensing Infrared and Terahertz Regions by Functional Films
177
T = 80 K c = 10 5 1
Ratio of dark currents
1.5
1.0
0.5
0.0
0
2 4 6 8 Average electron concentration, 1011 cm−2
10
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
Ratio of dark currents
1.5
T = 120 K 80 K 40 K
1.0
0.5
0.0
0
6 8 2 4 Average electron concentration, 1011 cm−2
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 ) / jth(QWIP ) as dark current than QWIPs do. Figs. 8.4 and 8.5 show the ratio jth a function of the average electron sheet concentration calculated using Equa-tions ( QW ) ( QD ) / p and different temperatures. The (8.6) and (8.10) for different ratios p −2 11 QD density is chosen to be ∑ QD = 5 × 10 cm with the maximum possible number
178
M. Willander et al.
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 ( QW ) ( QD ) p / p = 2c ∑ QD / ⎡2∑ QD − ∑ ⎤ . The quantity c can vary in a fairly ⎣ ⎦ 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 parameter 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.
j ph ≈
eG ph p
=
es
∑ p
I
.
(8.12)
Here G ph = s ∑ I is the rate of photoemission (per unit area), s is the photoemission 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
R
( QWIP )
=
es (QW )
∑
Ù p
( QW )
(8.13)
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 consideration is a rather weak (linear, near linear if the photoemission cross-section depends somehow on the electron concentration) function of the average electron concentration.
8
Sensing Infrared and Terahertz Regions by Functional Films
179
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 g (QWIP ) ≈
1 K p
( QW )
(8.14)
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 bottleneck 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 mechanisms of the electron injection from the emitter contact into the active region (tunneling 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 adjacent 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,
∑ − ∑D
≈
2C (V − VC ) e
(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
180
M. Willander et al.
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 experimental 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). D* =
H ph
(8.16)
2 ÙI H th
The QWIP detectivity can be expressed in terms of the average electron concentration
∑
by
D*(QWIP ) α
exp(
≈ s QW K
∑
s QW K
∑
∑
∑
exp(
ei ) 2k BT
) −1 (8.17)
T
exp( −
∑
2∑ T
) exp(
ei ). 2k BT
The detectivities of L-QRIPs and S-QRIPs are given by the following equations, respectively.
8
Sensing Infrared and Terahertz Regions by Functional Films
D*( L −QRIP ) α
∑
s QR K
exp(
(a 2 ∑ QR )1/ 4 exp(
≈
∑
s QR K (a
2
∑
1/ 4
QR
D*( S −QRIP ) α
)
exp(
s QR K (a
2
∑
∑ 1/ 4
QR
ei ) 2k BT
∑
∑
)
a
2
T
∑
exp(
) exp( QR
∑
p2
) −1
a 2 ∑ QR
∑ 2∑ T
181
(8.18)
ei ) 2k BT
2
16∑ T ∑ QR
) exp(
ei ). 2k BT
(8.19)
For the detectivities of L-QDIPs and S-QDIPs one obtains, respectively: D*( L −QDIP ) α
∑
s QD K
exp(
(a 2 ∑ QD )1/ 4 exp(
≈
s QD K (a
2
∑
∑ QD )
D*( S −QDIP ) α
1/ 4
exp(
s QD K (a
2
∑
2∑ T
∑ 1/ 4
QD
)
∑
∑
a
∑
p2 exp(
a 2 ∑ QD
T
∑ 2
ei ) 2k BT
) exp( QD
∑
) −1
(8.20)
ei ) 2k BT
2
16∑ T ∑ QD
) exp(
ei ). 2k BT
(8.21)
Relationships (8.17)–(8.21) show that the detectivity of detectors under consideration 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 detectivity 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): max D*(QWIP ) α 2 K s QW ∑ T exp(
ei − 1) 2k BT
max D*( L −QRIP ) α 2 K s QR (a 2 ∑ QR )1/ 4 ∑ T exp(
ei − 1) 2k BT
(8.22)
(8.23)
182
M. Willander et al.
max D*( S −QRIP ) ∝ s QR K
8∑ T ∑ QR
exp(
p (a 2 ∑ QR )1/ 4
ei 1 − ) 2k BT 2
max D*( L −QDIP ) ∝ 2s QD K ∑ T a 2 ∑ QD exp(
ei − 1). 2k BT
(8.24)
(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. Nevertheless, to estimate D*(S−QDIP) in such QDIPs, we set ∑ = 2∑ QD in Equation (8.21). As a result, we obtain: max D*( S −QDIP ) ∝ 2s QD K ∑ T ∑ QD exp(
ei ). 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 s QW max D*(QWIP ) max D*( L −QDIP ) s QD ≈ s QW max D*(QWIP )
(8.27)
a 2 QD .
(8.28)
Simultaneously, relationships (8.24) and (8.26) lead to: s QR 2 max D*( S −QRIP ) ≈ 0.8 (a ∑ QD )1/ 4 *( QWIP ) s QW max D
∑ ∑
QR
(8.29)
T
and s QD max D*( S −QDIP ) ≈ 2.7 *( QWIP ) s QW max D
∑ ∑
QD T
QD (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
8
Sensing Infrared and Terahertz Regions by Functional Films
183
(a)
T = 80 K
Normalized detectivity, a.u.
6
4
S-QDIP
2
0 T = 40 K
(b)
QWIP
6 L-QDIP 4 2 0
0
2
4
8
6
Average electron concentration,
1011
cm−2
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 −2 11 Fig. 8.6. We set for simplicity sQW = sQD. It is assumed that ∑ QD = 1 × 10 cm in 2 −2 11 an L-QDIP and ∑ QD = 5 × 10 cm (so the product a ∑ QD is approximately the 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
D ≈ e i exp( −
1 2
2 me i
∑
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.
184
M. Willander et al.
∑
QD
<
2 me i . e 2 ( In( i ))2 2k BT
(8.32)
Condition (8.32) can be rewritten as
∑ ∑
QD T
e 1 < 2p ( i ) 2 . k BT In (e i / 2k BT )
(8.33)
Using Equation (8.30) and condition (8.33), one can obtain the following inequality, s QD e i s QD max D*( S −QDIP ) 1 < 17( )( ) 2 =( )G *( QWIP ) s QW k BT In (e i / 2k BT ) s QW max D in which the right-hand side can be fairly large at not too small ratio
(8.34)
s QD
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 detectivity 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.
8
Sensing Infrared and Terahertz Regions by Functional Films
185
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 components 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 propagate 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 functional 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 impurities, which control the electrical properties of materials. Understanding the electronic structures of such impurities is very important in semiconductor physics and in applications of semiconductor devices. The impurities in bulk semiconductor 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
186
M. Willander et al.
heterostructures led to a revolution in semiconductor device applications. Therefore, the investigation of impurities confined in low-dimensional semiconductor 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 twodimensional 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
8
Sensing Infrared and Terahertz Regions by Functional Films
187
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 transition 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)).
Acceptor BE − + −
+
h-h coupling J=0,S=1/2 J=2,S=1/2
e-h coupling (Crystal-field scheme)
QW potential
J=1/2 (Γ6) 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
188
M. Willander et al.
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 tightbindinglike 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
H kin
)
L M 0 ⎤ ⎡P + Q ⎢ L+ P −Q 0 M ⎥⎥ =⎢ + ⎢ M 0 P −Q −L ⎥ ⎢ ⎥ + + M −L P + Q⎦ ⎣ 0
where P= Q=
g 12 2 k 2 m0
g 2 2 2 (k x + k y2 − 2 kz2 ) 2 m0
(8.35)
(8.36)
8
Sensing Infrared and Terahertz Regions by Functional Films
L = −i 3 M= 3
189
g 3 2 (k x − ik y )kz 2 m0
2 (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: QW H QW = DEhh ,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 coordinates 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 elements of the Hamiltonian can be computed analytically. Particularly, in the calculation 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 (r − z ) 2
2
= ∫ e − z q J 0 ( rq )dq.
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 experimentally obtained from optical spectroscopy such as infrared absorption,
190
M. Willander et al.
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 demonstrated 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 influence 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
Energy (meV)
20.00
2S-1S
2P-1S
15.00
10.00
5.00
Δ
Exciton
(lh-hh)
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
8
Sensing Infrared and Terahertz Regions by Functional Films
191
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 theoretical 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 possible 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 structures 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 electrontransit-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 example of a heterostructure with 2DEG, one can refer to a field-effect high-electronmobility 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 pronounced 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.
192
M. Willander et al. Gate layer (structure)
(a) Source
Gate
Drain W
+ + + + + + + + + + – – – – – – – – – – 2DEG channel Lg
Lc
Lc
High electric field gate-drain region
(b) Source
Gate
Drain
ΔLg Ld
(c)
2DEG channel
Gate layer
Gate
(d)
Gate RT structure
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
8
Sensing Infrared and Terahertz Regions by Functional Films
193
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 mechanism 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 heterostructures 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 electron transport model. The equations of this model comprise the Euler equation and the continuity equation: ∂u ∂u e ∂y + + nu = ⏐z = 0 ∂t ∂x m ∂x ∂∑ ∂t
+
∂∑ ∂x
=
J . e
(8.37)
(8.38)
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 introducing 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 ), k ∂z 2 ∂x 2
(8.39)
194
M. Willander et al.
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 conducting plane (gate) placed parallel to the channel at a sufficiently short distance 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 ). W k
(8.40)
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 ), respectively, 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 w (w + in ) =
4pe2 ∑ 0 W km
q2
(8.42)
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 spectra 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 damping or, vice versa, to its suppression. In some cases, the latter processes can result in a negative damping, that is, in plasma wave instability.
8
Sensing Infrared and Terahertz Regions by Functional Films
195
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: Wg =
4pe2 ∑ 0 W 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 conductivity 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 channels 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 heterostructure 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 fun0 damental plasma frequency, one obtains Wg / 2p ≈ THz . Because ∑ 0 in the gated 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).
196
M. Willander et al. 120
1.25 Lg = 0.5 μm = 0.75 μm = 1.0 μm
90
0.75 60 0.50
Quality factor
Plasma frequency, THz
ν = 2x1011s−1 1.00
30 0.25
0.00
0 0
0.2
0.4 0.6 Gate voltage swing, V
0.8
1
Fig. 8.11 Fundamental plasma frequency and quality factor versus applied gate voltage calculated 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
8
Sensing Infrared and Terahertz Regions by Functional Films
197
(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 resonant tunneling mechanisms are fairly strong, the responsivity of the detectors using former mechanisms can significantly exceed the responsivity of the resonant detectors 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 rectification) 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 responsivity R0 of the usual Schottky junction detector (without the excitation of plasma oscillations) can be expressed by Rw 1 ≈ 2 R0 ⎡⎣sinh (pn / 4Wg ) + cos2 (pw / 2Wg )⎤⎦
(a)
(8.44)
V0+Vω cos ωt + + + + + + + + + + + + + − − − − − − − − − − − − − Ohmic contact
(b)
Quasi-neutral 2DEG channel
V0+Vω cos ωt
Schottky contact
Depletion region
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)
M. Willander et al.
Responsivity, a.u.
198 μ = 30x104cm2 /Vs = 12x104cm2 /Vs = 6x104cm2 /Vs
1000
100
10
1 0.5
1
1.5 Frequency, THz
2
2.5
Fig. 8.13 Normalized responsivity as function of THz radiation frequency calculated for detectors 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 ≈ 2 2 R0 ⎡⎣sinh (pnw / 4Wu ) + cos2 (pw 2 / 2Wu2 )⎤⎦
(8.45)
where (compare with Equation (8.43)) Wu =
p 2 e2 ∑ 0 kmL
(8.46)
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 heterostructure devices such as detectors and frequency multipliers of THz radiation. However, the device proposals considered do not exhaust all interesting new ideas;
8
Sensing Infrared and Terahertz Regions by Functional Films
199
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 discovered 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 concept 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 background power load due to high sensitivity of the detector. Moreover, additional artificial 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
time
Cold-Electron Bolometer (CEB) P0 -removed by SIN junctions
Transition-Edge Sensor (TES) Ptotal = P0 + Pbias, Pbias = Pmax signal time
electron cooling!
(a) 0
Te cool
100 mk
Tph
Pbias- heating!
2
(b)
P0
230 mk
2
Te
0
P0
100 mk
Tph
Fig. 8.14 Comparison of the CEB and TES concepts (Kuzmin 2004)
230 mk
Te heat
Te
200
M. Willander et al.
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 electron temperatures lower than bath temperature with the corresponding improvement 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 antennacoupled 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
8
Sensing Infrared and Terahertz Regions by Functional Films
201
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 junctions. 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 ProtoGalaxy 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-
202
M. Willander et al.
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 = . (8.47) ∂P ∂Pw 4 −i w cv L + 5SLTe + ∂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 conductance 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 2 2 NEPtotal = NEPe2− ph + NEPSIN +
dI 2 . SI2
(8.48)
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 correlation term between these two processes: 2 NEPSIN = dPw2 − 2
dPw dIw dIw2 + 2 . SI SI
(8.50)
8
Sensing Infrared and Terahertz Regions by Functional Films
8x10
203
−18
NEP
NEP (W/Hz1/2 )
6
Goal: NEP=10−18 W Po= 0, R=6 k Ω 10−13 W, 6 k Ω 10−13 W, 0,5 k Ω
4
2
0 0.75
Po- microwave background load
1/2
δ ISQUID= 50 fA/Hz
0.80
0.85
0.90 V/Δ
0.95
1.00
1.05
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 bandwidth of 10% at frequencies in the range of 300–1000 GHz for background temperature 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
204
M. Willander et al.
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.
8
Sensing Infrared and Terahertz Regions by Functional Films
205
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 unrealistic 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 bolometer 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 NEP Bolo NEP Amp
NEP (10−18 w/Hz1/2)
5 4 3 2 1 0 100
frequency (Hz)
1000
Fig. 8.18 The dependence of the total, bolometer, and amplifier NEP of the cold-electron bolometer to the modulation frequency. We have stepped to the 19th power of CEB sensitivity for frequencies higher than 100 Hz. The dashed line represents the value of the NEPs as predicted by the nonequilibrium theory of the CEB
206
M. Willander et al.
Realization of the cold-electron bolometer would be the turning point from artificial 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 overcome 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
8
Sensing Infrared and Terahertz Regions by Functional Films
207
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
208
M. Willander et al.
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
8
Sensing Infrared and Terahertz Regions by Functional Films
209
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 Absorption Optical 14, 15 Actuator 2 Accuracy 2, 5, 8, 23 Aging 3, 8 Architecture Sensor 3, 4 Attributes Performance 4, 5
B Behavior plastic 12, 13 elastic 12
C Calibration 3, 6, 8, 9, 99, 106, 107, 120, 121, 122, 126, 133, 164 Ceramic 11, 14, 39, 70, 81, 149 Conductivity Thermal 11, 18, 93 Electrical 18, 82, 111 Ionic 114, 118, 120, 121, 168 Photo 190, 195 Confinement Geometrical 3, 11, 12, 14, 16 Quantum 169, 186, 187 Curve Calibration 3, 120, 121, 122, 133 Response 5, 23, 95, 96, 97, 122, 151
D Deposition 42, 45, 53, 54, 55, 58, 59, 60, 61, 62, 63, 70, 76, 102, 110, 127, 128, 132 Glancing-angle 32, 36, 48, 63, 64, 77, 78 Oblique angle 46, 47, 48, 50, 63 Material 2 Process 21, 31 Pulsed-laser 36 Vapor 31, 32, 33, 35, 36, 37, 45, 77, 78, 149, 152, 187 Electrochemical 32, 41, 63, 69, 77, 78 Thin film 32, 38, 46, 70, 76, 93, 100, 102, 112, 115 Nanostructure 77, 78 Sol-gel 151, 152 Detector Photo 94, 133, 170, 171, 172, 173, 174, 181, 183, 186, 208, 209, 210 Dielectric 10, 15, 16 Diffusion 7, 8, 21, 100, 127, 140, 153 Drift 5, 8, 28, 94, 101, 130, 132, 138
E Electronics 3, 5, 7, 9, 29, 80, 81, 82, 84, 96, 97, 100, 109, 134, 147, 161, 178 Energy 3, 6, 10, 11, 12, 13, 14, 15, 17, 18, 21, 27, 28, 36, 39, 68, 70, 71, 82, 88, 89, 90, 96, 101, 102, 106, 107, 111, 112, 113, 114, 123, 124, 129, 137, 170, 171, 173, 174, 176, 178, 185, 188, 189, 190, 191, 192, 193, 201, 206, 207 Surface 11, 13
211
212 Etching 2, 31, 32, 41, 42, 45, 70, 71, 72, 76, 78, 80, 83, 84, 165 Error Systematic 8 Random 8
F Fabrication 1, 2, 9, 15, 20, 24, 31, 32, 33, 34, 36, 38, 39, 44, 51, 55, 58, 59, 62, 63, 64, 65, 66, 69, 70, 71, 73, 74, 75, 77, 79, 80, 81, 82, 83, 84, 85, 87, 108, 114, 135, 137, 142, 144, 147, 151, 161, 162, 164, 165, 167, 171 Force Dispersion 11, 123 London 11 Van Der Waals 11 Frequency Resonance 2, 6
G Grains 12
H Hysteresis 5, 8, 127, 128 Homogeneity 13, 14
I Index Refractive 11, 16, 67, 99, 133, 134, 137, 150 Inhomogeneity 13, 14 Interactions Ionic 11 Interface 12, 61, 95, 133, 134, 137, 140, 172, 187
L Liquid 11, 12, 13, 14, 28, 32, 34, 38, 39, 41, 44, 67, 75, 89, 95, 105, 109, 110, 116, 120, 122, 123, 139, 167
M Materials Bulk 11, 15 Selection 1, 9, 16, 147 Measurand 3, 4, 5, 6, 7, 8, 9, 10, 17, 18, 19, 21, 22, 23, 24, 25, 26, 27, 28, 74, 198
Index MEMS 2, 16, 19, 21, 22, 24, 29, 68, 69, 74, 75, 76, 79, 81, 82, 84, 87, 88, 106, 129, 148, 149, 151, 152, 153, 161, 167 Metal 11, 12, 14, 15, 16, 21, 22, 36, 39, 44 Miniaturization 2, 106, 137, 138, 140, 142, 161, 167 Modulus Elastic/Young’s 11, 13, 16, 140, 151, 156, 159 Model 9, 12, 13, 14, 15, 16, 19, 39, 63, 132, 137, 138, 142, 144, 145, 176, 181, 190, 191, 192, 193, 195, 204 Multilayers 2, 98
N NanoStructure 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, 76, 77, 78, 79, 81, 83, 87, 129, 134, 137, 169, 170, 187 Technology 16, 43, 63, 64, 81, 82, 83, 85, 208 Scale 11, 12, 13, 14, 15, 60, 61, 62, 63, 80, 81, 82, 85, 137, 140, 144 Materials 11, 12, 13, 14, 63, 81, 114, 129, 134 Morphology 7 Fluid 13, 14, 16 Instrument 10 Particle 11, 15, 16, 27, 31, 32, 41, 44, 45, 62, 76, 77, 78, 80, 81, 107, 134, 135, 139, 140, 144 Regime 11, 13 Noise 3, 6, 7, 8, 10, 18, 24, 94, 95, 96, 110, 111, 119, 129, 135, 137, 138, 150, 161, 187, 202, 204, 205, 206, 207
P Point Melting 11, 12, 16 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, 112, 114, 116, 119, 123, 124, 125, 126, 127, 129, 130, 131, 132, 134, 135, 138, 139, 140, 141, 142, 143, 144, 145, 150, 153, 154 Precision 2, 5, 8, 24, 29, 100, 102, 137, 161
Index Process Fabrication 2, 9, 20, 32, 51, 62, 66, 74, 147, 161, 162, 164, 165 Properties Physical 132, 145 Chemical 2, 11 Mechanical 11, 127 Optical 14, 15, 16, 62, 63, 64, 134, 135, 139
Q Quantitative Sensor 4, 5, 132
R Range Dynamic 5, 21, 22, 108, 109, 149, 150, 201, 208 Reliability 5, 10, 15, 108 Repeatability 8 Resolution 5, 7, 8, 24, 66, 67, 68, 70, 80, 81, 84, 108, 134, 135, 137 Rheology 10
S Scaling 2, 8, 9, 10, 11, 13, 15, 16, 28, 31, 32, 47, 50, 58, 69, 83, 110, 112, 132, 143, 144, 162 Selectivity 2, 5, 8, 9, 71, 106, 108, 109, 110, 116, 117, 123, 129, 138, 165 Self-assembly 2, 32, 64, 76, 77, 78, 80, 106 Semiconductor 27, 29, 41, 62, 63, 64, 65, 70, 71, 72, 74, 75, 79, 82, 93, 101, 114, 129, 142, 147, 150, 152, 169, 170, 187, 188, 193, 208, 209, 211 Sensitivity 5, 6, 7, 8, 9, 13, 15, 18, 21, 22, 23, 28, 60, 88, 91, 92, 93, 94, 95, 108, 109, 112, 116, 119, 125, 126, 127, 129, 130, 135, 137, 143, 150, 167, 171, 182, 201, 202, 204, 207, 208 Signal-to-noise ratio 95, 97, 137, 138, 187 Silicon 2, 19, 20, 21, 22, 23, 25, 26, 28, 70, 71, 72, 75, 76, 77, 80, 83, 99, 150, 152, 153, 162, 163, 164, 165, 167, 187 Solid 12, 13, 29, 34, 36, 39, 41, 62, 63, 72, 81, 82, 83, 98, 99, 107, 116, 134, 140, 141, 142, 144, 167, 169, 192
213 Spectrum 3, 15, 18, 26, 89, 90, 91, 96, 101, 102, 103, 173, 174, 175, 196, 197 Stress 8, 18, 19, 20, 21, 24, 27, 29, 65, 74, 147, 148, 150, 152, 154, 157, 158, 159, 160, 163, 164, 166, 167, 190, 193 Surface 2, 3, 11, 12, 13, 14, 19, 21, 24, 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, 107, 112, 113, 125, 130, 132, 134, 135, 138, 140, 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, 81, 82, 83, 88, 89, 93, 94, 95, 96, 97, 100, 101, 103, 105, 106, 107, 109, 110, 120, 126, 127, 132, 134, 135, 137, 138, 139, 140, 142, 143, 144, 145, 149, 151, 161, 162, 167, 174, 188, 190, 193, 195, 196, 197, 201, 202, 204, 205, 207, 208
T Temperature glass-transition 11, 12, 72, 116 Tension Surface 12, 41 Thin films Functional 2, 3, 13, 76, 78, 87, 148, 150, 161, 166 Time Response 5, 7, 25, 93, 94, 95, 101, 103, 127 Threshold Sensor 4 Transducer 1, 2, 4, 5, 6, 7, 8, 9, 10, 14, 15, 16, 17, 18, 22, 65, 66, 67, 70, 73, 74, 75, 76, 77, 79, 80, 105, 106, 107, 108, 110, 111, 112, 114, 123, 127, 129, 137, 138, 141, 142, 144, 145, 166 Transduction 3, 6, 9, 10, 15, 17, 18, 19, 21, 22, 24, 25, 27, 28, 105, 106, 107, 109, 110, 113, 114, 116, 123, 129, 132, 138, 147, 148
V Variability 2, 8, 97 Van der Waals 11 Viscosity 11, 13, 14, 16, 18, 39, 41, 44, 195