SELECTED SENSOR APPLICATIONS IN MANUFACTURING
Practitioners of manufacturing have always employed sensors. The earliest sensors included the experienced eyes of a blacksmith helping to guide the shaping of a steel sword and the vernier calipers of a medieval artisan used to measure the critical dimensions of an architectural icon. As the technological sophistication of the manufactured products has increased, so has the complexity of the process required for manufacture. To a significant extent, manufacturing has been in the fore-front of incorporating advanced sensor technology. For example, the current revolution in computer-integrated manufacturing has been enabled by advanced sensor technology.
An increasingly competitive global market-place has demonstrated the high cost of "inspecting-in" quality (i.e., the cost to assess the quality of the product after it has been manufactured and then "fix" or remake those parts that do not meet the required quality level). Consequently, quality must be an integral element of the manufacturing process, necessitating that the process be under control, either through constant monitoring using appropriate sensors or by withdrawing the product for inspection at intermediate manufacturing stages. In many instances today, manufacturing process sensors are the limiting capability that defines the best possible product performance and reliability.
The traditional manufacturing approach involved calibrating the equipment at the start of the operation with the expectation that it would continue to perform satisfactorily over a certain period of time. In this approach, sensors monitor process parameters, such as temperature, gas pressure, and composition. However, as the sophistication and complexity of materials processing increase, sensors will be needed to directly monitor changes in the product, such as grain size growth during processing; location of nucleation sites in epitaxial grown thin films; and chemical composition, morphology, and nanoscale thickness of fiber interface coatings for ceramic and metal matrix composites. In addition, on-line control of processes is highly desirable when the use and generation of toxic or hazardous chemicals is involved. In some cases, the component may be so sophisticated that manufacturing it reproducibly would not be possible without in situ sensors to provide on-line measurement and feedback for real-time process control.
An understanding of the process based on first principles or empirical study is highly desirable for manufacturing complex components. This approach provides an understanding of the interdependence between the various processing steps and can result in the development of qualitative and quantitative models for in situ model-based control. For example, variations in the product that would be caused by variations in a given process parameter, such as temperature, can be calculated; the process could then be adjusted to accommodate these variations. Furthermore, the combination of process models and sensor-based on-line control permits downstream processing steps to be tailored to accommodate irregularities in the current (or preceding)
step. Such "feed-forward" control optimizes the process window for each individual process step, resulting in intelligent materials processing of a high-quality product at high-yield production rates.
Many of the recent advances in control technology have been made possible by rapid progress in information processing. Real-time control is only possible when the time constant for the measurement and analysis is commensurate with the time scale of the process itself. In the past, process monitors were necessarily confined to measurements of primary parameters, such as temperature and pressure. The massive computing power afforded by modern computer workstations now allow in situ, real time measurement (assuming the appropriate sensor technology is made available) of parameters that were not feasible in the past. With computational performance doubling about every two years, even more precise real-time control will become the standard practice of the future.
As is evident from the preceding discussion, sensors are crucial in a wide variety of manufacturing situations. Sensors that cost-effectively measure critical material behavior and guide a process to achieve desired properties would greatly enhance the process productivity and yield. For manufacturing, this is often the most important, yet most elusive, process improvement.
The remainder of this chapter extracts examples of current needs in sensor materials by examining needs from the high end of manufacturing: intelligent processing of materials and manufacturing of products that require a multitude of interdependent steps. The next section discusses sensor needs that arise from the self-directed curing of high-value polymeric composites. The following section broadly surveys the manufacture of optoelectronic components (integrated circuit and optoelectronic devices) in order to identify key sensor needs. The combination of these examples then serve to define key research objectives for sensor materials, which are summarized in the final section.
INTELLIGENT PROCESSING OF ADVANCED MATERIALS
Different terminology has been used to describe intelligent processing, including closed-loop (involving sensor feedback) and real-time requiring dynamic adjustment of control variables. Although closed-loop and real-time are necessary, intelligent processing also requires knowledge about the material and the process to enable a self-directed control system. In short, a self-directed system generates a control path in response to changes in material behavior that are denoted as process events, as opposed to some pre-established schedule of process parameters. Intelligent processing systems exemplify a process control strategy whereby process events regarding material changes, such as chemical-state change, flow, deformation, and growth, are continually evaluated for dynamic adjustment and prediction of process parameters that affect product quality metrics, which include process repeatability, process yield, and consistent properties.
Discrete part production, as opposed to the coordinated control of multiprocess manufacturing, is typically the domain of intelligent processing. Processes such as machining, welding, and forging typically involve either a priori or a posteriori control strategies. Although the metrics are very similar, intelligent processing is distinguished from conventional a priori and a posteriori processing by the in situ control system autonomy used in achieving these metrics, as depicted in Figure 3-1.
In a priori processing systems, control actions for variables such as process temperature and pressure are based on the results of a model (that may have been developed by empirical trials), which are used to generate a predicted schedule of process
events; e.g., a time-temperature profile. In a posteriori statistical process control, attention is focused on monitoring a process to collect statistics. These statistics are used for process evaluation in order to eliminate or reduce undesirable variability. Reduction of process variability is accomplished by modifying either the process or those product features that are the causes of the variability.
Intelligent processing systems depart from predefined process models and statistics by using sensed information to self-direct the process via an event-based control strategy. Event-based implies the use of sensed information to denote the occurrence or prediction of an event for which adaptation (i.e., on-line decision making or conflict resolution) of some control variable may be required. The use of sensed information, such as in situ material behavior, to effect continual adaptation of these variables is the basis for the term self-directed as it is applied to intelligent processing (LeClair, 1991). Self-directed control can be used in conjunction with process models or in lieu of them if sufficient empirical information exists to construct the necessary rule-bases or neural networks.1 In summary, intelligent processing is process control by objectives rather than control by following prescribed parameters.
Applications of intelligent processing vary widely and range from the processing of thin-film engineered materials2 to the processing of polymers and polymeric composites, metals, and ceramics for bulk structural applications. In almost every application to date, intelligent processing has been applied to introduce a new material or process or to achieve a step-change improvement in an existing process that has been plagued by quality problems.
General Sensor Issues
A key issue raised in past reports of the National Materials Advisory Board (as referenced in the Preface) is that sensors are the weak link in intelligent processing. There are fundamental concerns about the capability of other available sensors to perform in a noisy or adverse manufacturing environment, compounded by real-world constraints that include:
short time constants: sensing very rapid, localized changes or large gradients in a material over several hours of processing;
chemical change monitoring: sensing the nonlinear behavior of one or more simultaneous chemical reactions and associated by-products;
point measurement: sensing critical points, rates, and changes in rates within a complex three-dimensional product shape; and
inferred measurement: sensing parameters of interest that are not directly measurable (i.e., inferred parameters require the establishment of constitutive relationships to those parameters that can be directly measured).
Intelligent processing of polymeric matrix composites offers a clear illustration of the challenges posed by these issues and benefits that could result from advances in sensor materials and technology.
Sensors for Intelligent Processing of Structural Polymeric Composites
Structural polymeric composites are critically important to sustaining U.S. aerospace and defense superiority. The potential market for these materials, by the end of the century, is projected to be 90,000 metric tons, a sixfold increase in tonnage since 1986, with a tenfold increase in worldwide employment to 200,000 people (AIA, 1991). Improvements in the manufacture of products that use these materials will help in the realization of such projections.
A polymeric composite typically has two primary microstructural components: a polymer material and reinforcing fibers. Two different generic types of polymers are used: thermosetting (the traditional choice) and thermoplastic (the more recent choice). Thermosetting polymers are initially low-viscosity liquids that can flow into a mold or around fibers. During the course of processing, thermosets react to increase molecular weight and viscosity, eventually becoming highly cross-linked, insoluble, infusible materials. On the other hand, thermoplastics are fully polymerized materials that melt and flow upon application of heat. Thermoplastics are processed well above their glass transition temperatures or melting points (if the material
is semicrystalline) to reduce the melt viscosity and allow flow and to promote adhesion to the fibers.
The fibers in the composite are the component that provides the desired high stiffness and strength properties that make the material useful for structural applications. The polymer matrix protects the fiber, serves as a medium to transfer load between the fibers, and stabilizes the fiber when subjected to compression loading.
A composite structure is typically made by stacking layers of pre-impregnated ("prepreg") material3 in prescribed directions on a tool form, compressing the stack, and then curing or consolidating the composite. In the case of high-performance thermosetting resins, the curing process requires the application of heat and pressure. For small-lot production, this curing step is usually accomplished in a large pressure vessel called an autoclave. Alternative processes, such as resin transfer molding, are preferred for higher-volume mass production.
Consolidation of a thermoplastic resin does not require autoclave but does require the application of energy, such as heat. In addition, new composites are being developed that consist of blends of thermosets and thermoplastics; in these materials, curing and consolidation occur simultaneously and interactively to produce complex microstructures possessing functionally gradient material properties that can be tailored to optimize specific strength, processability, etc.
Intelligent processing can improve the resultant properties of composite components. The discussion that follows examines the technologies required to fully implement intelligent processing with emphasis on the most critical sensor requirements.
Thermosetting Polymer Matrix Composites
At the present time, most of the structural polymeric composites used in the aerospace industry consist of a high strength/high stiffness fiber (e.g., graphite, boron, or aramid) embedded in a thermosetting organic-resin matrix binder (e.g., epoxy, bismaleimide, polyimide, polyester, or cyanate ester). These resins are complex chemical formulations with batch-to-batch chemical variations.
Until recently, the conventional processing strategy for thermoset polymer composites was based on a priori models. These models cannot account for variations in material chemistry, part geometry, autoclave sizes and heating patterns, and tolling materials. As a consequence, they assume "one size fits all" and are developed for the worst-case processing conditions. Hence these models specify a safe (i.e., protracted) cure cycle.
In contrast, intelligent processing senses the state of the material during cure. The development of these models requires the definition of events that denote the changes in material state, e.g., flow, deformation, growth, etc. A processing system is also required that is empowered (through a model or set of condition-action relations) to adjust the process parameters in response to changes in the materials state to achieve desired end-use properties.
The first steps in developing an intelligent processing system are the mapping of these desired end-use properties to the actual material parameters that can be sensed and the processing variables that can be controlled. Some properties (such as residual stresses; void volume; and fiber location, volume, and orientation) can be inferred from engineered relationships with parameters that can be sensed in situ, while other properties (such as part strength and modulus) depend upon other engineering relationships to determine properties that cannot now be measured in situ.
Identifying "need-to-be-sensed" parameters can be accomplished through a review of available models of the process4 or by developing a mapping of properties to parameters and to variables. This mapping, depicted in Figure 3-2, is often referred to as an "influence diagram." It relates process control variables to in situ material behavior to resultant material properties.5 It can serve to identify and prioritize relationships needed for self-directed process control as well as establish sensing technology requirements (LeClair and Abrams, 1989).
In the next step, a process engineer must assess available sensor technologies that can provide the necessary information either directly or indirectly. To accomplish this assessment, the framework developed in Chapter 2 can be used as a guide to list and detail the necessary parameters for evaluating
and comparing different sensor technologies. With these descriptors, the process engineer can describe the range, tolerance, and limits of the sensing requirements and identify one or more candidate sensor technologies, and their respective capability and constraints.
Performance estimates and benefits of intelligent manufacturing can only be validated in a true manufacturing environment. One example is provided by the U.S. Air Force for an application that produces polymer composite replacement parts for the A-10 aircraft. In this case, sheet metal aluminum parts for the leading edge of the aircraft wing are being replaced with a hybrid carbon and aramid fiber/epoxy-resin composite to improve performance and extend the useful life of the aircraft. Initially the composite replacements were cured using a conventional (i.e., a priori model) cure cycle. But in 1990, the Air Force's Sacramento Air Logistic Center implemented an intelligent processing system that achieved full cure in 1.5 hours versus 7 hours required for the conventional approach; that is, a 70 percent reduction in cure time. The intelligent manufacturing system was originally implemented using the polymeric material temperature (as measured by a thermocouple) as the basis for estimating the degree of cure. Figure 3-3 displays the results for the initial implementation. Autoclave temperature profiles during cure are presented for both approaches along with the resultant test values of selected material properties. This data confirmed that intelligent processing could produce parts with strength values statistically equivalent to those conventionally processed (Warnock and LeClair, 1992).
Even though this initial implementation was quite successful, it was not without difficulty. For instance, the thermocouple signal was "noisy" and hence many data points were needed to clearly establish the trendline. To identify various levels of candidate improvements, the framework presented in Chapter 2 was applied to relate sensing needs (e.g., resin temperature) to the currently available sensor technology (e.g., thermocouple); the results of this comparison are summarized in Appendix D. Some of the thermocouple shortfalls can be partially ameliorated, but nonetheless, a new sensor technology was clearly needed. For example, autoclave temperature does not directly relate to the degree of resin cure (i.e., extent of the formation of cross-linked polymer bonds).
The new sensor technology should be able to
measure microscopic properties (e.g., physical, mechanical, chemical) in situ while minimizing the number of different sensors required for intelligent processing. Ideally, a new sensor technology should be as convenient, practical, and inexpensive to implement as a thermocouple.
One such multiuse technology which addresses both the limitation of thermocouples and the need for more microscopic property monitoring is photon scattering fiberoptic sensing, as described in Appendix D. Photon-scattering sensor technology6 uses the energy and momentum distribution in the scattered photon flux to extract a wealth of useful and interpretable information regarding the physical and chemical nature of the material being processed. Suitable electromagnetic radiation scattering measurements (i.e., Rayleigh, Brillouin, and Raman scattering) can theoretically provide a direct measurement of material properties that include bulk modulus, thermal diffusivity, sound attenuation factor, sonic speed, heat capacity ratio, chemical composition, bulk viscosity, shear viscosity, and the free energy of mixing (Maguire and Talley, 1995). The scattered flux can be detected by the receiving optical fiber at any given angle, and transported at the speed of light to a remote spectrometer.
Near-infrared Raman spectroscopy, using fiber optics to provide a relatively simple optical contact with the material, can allow direct determination of the degree of cure. As an example, an epoxy-amine material7 has a Raman spectral peak at 1,250 cm-1 resulting from the vibration of the polymer's epoxide ring structure and another one at 2,870 cm-1 arising from the stretching of C-H bonds. During cure, the number of the C-H bonds does not change significantly, while the number of epoxide groups notably decreases. Therefore, the 2,870 cm-1 peak can be used as an internal standard to determine the degree of change in the 1,250 cm-1 peak over time (Maguire and Talley, 1995). In this case, the ratio of the two peaks at time, t, can be represented as:
The degree of cure, a, can be defined as:
This parameter, a, is a measure of the degree of cure related to a direct material phenomenon (Maguire and Talley, 1995).
For example, Figure 3-4 compares a near infrared fiberoptic Raman spectrograph for an uncured and cured sample of an epoxy-amine. From this spectrum it is apparent that R(t=0) is about 0.7. After five hours of curing, R(t=5 hrs) has fallen to about 0.3.
Through a series of such measurements, the state of cure can quickly be determined. As indicated in Appendix D, Table D-3, expansion of the photon-scattering fiberoptic sensing into the Rayleigh and Brillouin regimes can provide additional useful information and could serve as the basis for an advanced noncontacting multifunctional
sensor array useful in many intelligent processing applications. Continued research and development of multiuse sensor technology will provide crucial enabling science and understanding for addressing these intelligent processing sensor opportunities.
The last section of this chapter discusses research needs and opportunities for photon-scattering sensor technology.
Thermoplastic Polymer Matrix Composites
Future systems, ranging from aircraft parts and printed circuit boards to applications that must survive in the rigors of space, will require further innovation in intelligent processing, particularly in the development of low-cost out-of-autoclave8 composite fabrication techniques and new material systems such as components made of thermoplastic matrix composites. A newly developed out-of-autoclave technique involves the use of a unique energy transduction mode known as direct electric heating, that is, passing a current through the carbon reinforcing fibers (Miller and Van den Nieuwenhuizen, 1993). Electric heating of the fibers provides for a versatile temperature source capable of accommodating complex parts possessing varying curvature and laydown angles. It can produce a high strength, low void content composite. In the discussion that follows, processing needs, together with the unique hardware for part lay-up, are discussed to identify requirements for new sensor configurations.
Direct electric resistance heating uses electric current to heat the graphite fiber prepreg near the laydown point while the material is in the form of prepreg tape that is fed to a tape head that applies pressure and electrical power via metallic pads. A
schematic of this approach is depicted in Figure 3-5. The process requirement consists of raising the temperature of the high-performance thermoplastic matrix at the hot zone or interface between the part and tape.9 The forming of the thermoplastic is achieved by pressing the part and tape together for continuous "wetting" contact and allowing the interface to cool so that the tape adheres to the part (Anderson and Grant, 1991; Cirino et al., 1991).
Thermoplastic material part quality is defined by the resultant physical properties, such as ultimate strength, void content, and elastic modulus. However, in-process measurements of some of these quantities (e.g., ultimate strength) cannot be directly obtained, since their measurement requires part destruction (University of Delaware, 1989).
Void content, a commonly used measure of laminate quality, can be measured by x-rays, ultrasonic pulses, or determination of local electrical or by thermal conductivity using infrared sensors. Controlling void content during the process requires measuring both temperature and pressure in real-time during processing and making appropriate adjustments. Thus, temperature must be measured; if it begins to exceed the prescribed range, the tape head pressure can be adjusted accordingly across the face of the tape head to bring the temperature back in line.
The issues and resulting sensor development opportunities as a result of these needs are summarized in the last section of this chapter.
SENSORS FOR ELECTRONICS MANUFACTURING
Electronics has become the largest single industry in the world, with a projected worldwide market of $2,000 billion by the year 2000, including U.S. revenue in excess of $400 billion and a job base in excess of 2.9 million employees. Integrated circuits (ICs) based on semiconductor materials are the enabling technology for the diverse range of functions, such as logic, memory, and control, that are required for today's electronic products. The national economy and infrastructure, including the manufacturing and service sectors, rely to a large degree on the information processing afforded by ICs. The future of the U.S. economy and its national security are directly linked to the health of the domestic microelectronics and electronics industry.
In addition to the rapid growth of silicon-based IC industry, advances in the understanding of compound semiconductors, typified by gallium arsenide (GaAs), have led to major progress in optoelectronics. Advances in materials control offered by new growth techniques and increasingly precise sensors have enabled the development of new types of "band-structure-engineered" materials that are leading to major new areas of optoelectronics as they are incorporated into devices. These devices range from laser diodes to optical fibers (which form the basis of optical communication) to fiber-based nonlinear optics, lasers, and optical amplifiers pumped by strained quantum-well lasers. The marriage of optoelectronics and microelectronics could lead to optoelectronic integrated circuits that enable such advances as massively parallel optical interconnects for high-speed computing and communications applications.
Integrated Circuit Manufacturing
The increasing complexity of microelectronics has caused a dramatic escalation in the cost of semiconductor production facilities. New production facilities for volume manufacture now cost approximately $1 billion. This cost is driven by the complexity of today's ICs and the sophistication of the processing equipment required for their manufacture. As the technology advances, the costs for future factories to produce the next generations of
technology are anticipated to be even greater. Because of the major capital investment required to survive in this highly competitive industry, new approaches are required to increase the productivity of semiconductor factories. These approaches will make extensive use of sensor technologies to provide increased understanding and control of the various manufacturing processes.
Without extremely stable and well-characterized processes, it is not possible to obtain an acceptable final product without employing either sensors that monitor the process steps along the way or sensors to measure the state of the product at intermediate phases of the manufacturing sequence. The advancement of semiconductor manufacturing toward more and more complex structures, requiring smaller and smaller feature sizes, is rapidly leading to the convergence of the two cases.
A few observations serve to illustrate the complexity of IC manufacturing. Manufacture of a typical IC, such as a 16-megabyte dynamic random-access memory chip, requires in excess of 400 distinct process steps involving different types of materials. Typical process steps include ion implantation and annealing to control the electrical properties of the semiconductor, metal deposition for electrical contacts, metal patterning based on lithography and etching for interconnects, and dielectric deposition and patterning for insulation between the layers of multilevel metal interconnect required to carry the signals from the individual electrical components in the circuit. The inexorable march to higher and higher density requiring smaller and smaller feature sizes has advanced to the point that today's circuits have feature sizes below 1 µm; feature sizes for advanced devices and circuits are currently 0.35 µm.
The manufacturing complexity is typified by the fact that an IC contains several layers of these 0.35-µm interconnects separated by insulators, and these layers must be aligned within a small fraction of the 0.35-µm line width. Further, to achieve the performance required, the chemical compositions of all parts must be controlled with enormous precision, and the cleanliness of the process maintained at unparalleled levels. For example, a particle with a diameter of about one-tenth the minimum feature size can ruin the entire IC. Since the size of a typical IC is 2 cm × 2 cm, this implies that sensors must be developed to detect less than one 35 nm particle per 4 cm2.
Another consequence of the steady advance of IC technology to smaller feature size is that the composition of the semiconductor must be controlled over a 10-nm scale and the composition of the dielectrics on a 5-nm scale. This control will require increases in the understanding of materials science and control of the process steps to produce the necessary feature sizes with required absolute interface and compositional control. It is also evident that the complexity of structural control for IC manufacturing will approach that for artificially structured materials discussed below.
Low-cost, sensitive, and reliable sensors are required to increase the rate of learning in process tool development, reduce the time to market for process equipment, improve tool and process control, improve process yield, and reduce defects (SIA, 1994). Increased use of real-time in situ sensors is driven by economics for these applications. Sensors are the critical elements in closed-loop process control and are necessary for detecting process problems when they occur, so that corrective action can be taken immediately. Sensors are also required to improve first-pass success when introducing process variations. For example, accurate control of even such commonly used processes as rapid thermal processing and plasma deposition and etch requires new sensor materials and approaches. And environmentally conscious manufacturing will require recycling and reuse of chemicals, not only for waste minimization but also for cost reduction. Typical sensor needs for these applications are discussed in the last section of this chapter.
Processing of Artificially Structured Semiconductors
The complexity of the semiconductor systems that are spawning development of artificially structured compound semiconductors requires major improvements to sensor materials and technologies for volume manufacture. In many cases, the technology requires that the microstructures be tailored at the atomic level in order to attain precise control of the resultant electronic properties.
The control required for low-cost volume manufacture of the artificially structured materials, such as those that form the basis of the LWIR detectors (which are themselves sensors, discussed in Chapter 5), necessitate the development of in situ sensors for intelligent on-line process control. Small numbers of devices with the desired performance can be demonstrated in a research environment. However, without real-time control the process cannot be scaled up at high yield to produce devices in which the composition is consistently controlled to a fraction of a percent, the dopant concentration controlled to parts per billion, and the thickness controlled to an atomic layer.
Techniques for characterizing epitaxial device structures have long provided crucial feedback to crystal growers as they try to develop and optimize growth processes. Techniques such as x-ray diffraction, secondary-ion mass spectrometry, reflection ion and optical spectroscopy, and photoluminescence permit one to determine thickness, composition, and quality of even very complicated multi-layered structures quickly and precisely. Many of these techniques have conventionally been performed off-line.
Over the past few years, there has been a considerable effort to bring as many as possible of these characterization techniques into the crystal-growth reactor itself. The motivation has been two-fold: to gain an increased understanding of the complex science of crystal growth and to improve the precision, quality, and technology of crystal growth for manufacturing.
Improvements in crystal growth of semiconductor materials will require the development of a full panoply of surface-sensitive structural, chemical, and optical diagnostic sensors. These sensors must be compatible with contamination-free manufacturing environments that include rotating wafers and harsh chemicals. Among the more promising of these diagnostics are optical transmission spectroscopy (the basis for a recent technique for measuring wafer temperature) and reflectance spectroscopy (recently applied to the growth of complex vertical-cavity surface-emitting laser structures).
There is currently a limited understanding of the relationship between processing parameters and the final product. As a result, process control requires measurement of both the chemical composition of the gases in the growth chamber and the composition and thickness of the material being deposited. The development and application of new types of sensors will lead to increased understanding of the relationship between the gas phase and the resulting solid to permit the development of algorithms for on-line control. R&D of in situ diagnostics is being pursued intensively in many laboratories throughout the world. Advances in techniques and development of new techniques will undoubtedly yield important advances in manufacturing technology over the next several years.
Noteworthy among applications are molecular beam epitaxy and chemical vapor deposition processes for epitaxial growth of electro-optical thin-films for semiconductors, detectors, etc. Spurred by the growth in new high-bandwidth wireless and optical fiber communication, advanced semiconductor processing has become a very active area of intelligent processing research because of the potential size of the worldwide market. These epitaxial growth processes require sensors for control of layer thickness, alloy concentration, interface sharpness, composition, etc., to enable low-cost, reproducible, uniform, and tailorable structures. Research in new sensor materials and technologies for molecular beam epitaxy and chemical vapor deposition includes work on reflection mass spectrometry (Brennan et al., 1992; Chalmers and Killeen, 1993; Chalmers et al., 1993); desorption mass spectrometry (Evans et al., 1993a, b), and ellipsometry (Patterson et al., 1992). The focus of these research efforts has been in developing non-invasive sensor technologies to sense and control the thickness (down to one atomic layer) of films, especially ''superlattice" structures; a key sensor materials issue arising from the need for an optoelectronic modulator is discussed in the last section of this chapter.
SENSOR MATERIALS NEEDS IN MANUFACTURING
Curing of Thermosetting Resins
As previously discussed, a specific sensor suite for monitoring the curing of thermosetting resins has
been developed but has several shortcomings. The principal opportunities to improve sensor materials and technologies include measuring more microscopic properties while reducing the number of separate sensors required for intelligent processing, and increasing the range of materials that a sensor technology can accommodate. Multiuse sensors provide an important method for addressing these intelligent processing sensor opportunities.
Photon-scattering intrinsic sensor technology (described in Appendix D) has the potential to satisfy many of these needs cost-effectively. The near-term research opportunity is to use photon-scattering sensor technology to increase the range of properties that laser-fiber optics can accommodate for multiproperty sensing (e.g., resin temperature, viscosity, degree of cure, degree of surface stress). The advantages of this technology include small size, ruggedness, survivability in hostile and inaccessible environments, simplicity of construction, ability to monitor multiple parts at one time in an autoclave, and low cost.
Laser-fiberoptic sensor technology has been demonstrated to perform in situ chemical analysis of polyimide and epoxy to determine degree-of-cure and infer the molecular weight during processing (Maguire et al., 1992). This accomplishment in laser-fiberoptic sensing technology has enabled the intelligent processing of new polyimide composites that are more difficult to process and for which conventional sensing of physical parameters such as temperature and viscosity has been insufficient. Since laser-based fiberoptic sensing has great potential, a comparison of this sensor technology against the multiproperty sensing requirement is contained in Appendix D.
Just ten years ago, the ability to use photon scattering as the basis for an industrial sensor-based control technology would not have been possible, and it is noteworthy that the current use of this technology is due to advances in four related areas:
the invention of robust low-cost lasers to provide a ready source of photons;
low-loss optical fiber, developed for the telecommunication industries, to allow the transport of photons over exceedingly long distances;
the development of sensitive strained superlattice infrared detectors (discussed in Chapter 5) to enable the detection of the scattered "heat" flux with unprecedented sensitivity; and
the massive increase in computational capability that has taken place over the last ten years to allow these powerful individual technologies to be suitably combined.
Consolidation of Thermoplastic Resins
For intelligent processing of thermoplastic structural composite parts, void content sensing is currently the most common measure of laminate quality. Void content can be measured by a number of techniques, including x-rays, ultrasonic pulses, and determination of local electrical or thermal conductivity using infrared sensors. The choice of a technique for in situ process monitoring is governed by factors such as speed of response, cost, and reliability. The use of x-ray techniques requires complex, costly, shielded equipment and a diagnostic system that must compute the control response within 100 milliseconds. The development of such a system would be a substantial undertaking.
For ultrasonic sensors, the main issues are essentially those of implementation: that is, how to couple the energy produced by the ultrasonic transducer into the newly laid-down hot tape and how to detect the reflected signal. A suitable cooled, miniature transducer would be required; this would necessitate modification of existing transducer designs. The detector required to receive the reflected signal information would involve the development of a sensor array, suitable low-attenuation coupling greases or gels (which would not interfere with laying down the next layer of prepreg), and a fast multiple-channel signal processor. For this sensing solution, the limitations imposed by transduction mechanics essentially define the operating parameters.
The use of thermal conductivity measurements with an infrared scanner is an attractive option. The presence of voids changes the rate of cooling of the newly laid-down tape and a thermal image of the part shows the void areas as hot spots. The sensor does not need to contact the part. Optical access to the part is required, however. A commercially available scanning thermal imager ("pushbroom"
array) can be used. The primary implementation issues are optical access and processing speed. An optical fiber that conducts suitable infrared energy and could be built into or attached to the tape-laying head, with perhaps direct fiber bonding to a detector, is required. For this sensing solution, process geometrical constraints would be of major importance.
In order to control void content in real-time during the process, both temperature and pressure must be controlled. Thus, temperature must be accurately measured, and if too high, pressure must be adjusted across the face of the tape head. This leads to a requirement for local point measurements of pressure, implying an array of transducters at pitch spacing of less than 2.54 mm (0.1 inch). Conventional technology allows this to be done with an array of pressure tubes attached to miniature pressure transducers. But this process is bulky, the time constant is too slow (due to tube gas path), and tube blockage by molten matrix material is likely. A pressure-conductive matrix pad, which backs the metallic electrical pressure shoe, is geometrically compact and digitally simple (computationally), but it is thermally limited in currently available versions.
The measurement of modulus as a control parameter is currently not possible, due to the non-isotropic nature of the material and the thermal gradients in the parts. The best indirect approach is to measure the speed of sound (which is directly related to modulus). Other measurement concepts, such as the determination of molecular bond strengths of the matrix material at the interface, appear possible but are of unknown feasibility. Thus there is a requirement for the development of a matrix modulus sensor.
Manufacturing Integrated Circuits
Accurate control of even such commonly used processes as rapid thermal processing and plasma deposition and etch requires new sensor materials and approaches. For example, the back surface emission technique for measuring and controlling temperature in today's rapid thermal processing equipment leads to error of as much as 50¹C to 200¹C in temperature. The new long-wavelength infrared sensors, discussed in Chapter 5, could be important for more accurate measurement and control of this temperature.
Improved sensors that monitor gas and chemical purity/cleanliness are of major interest. Gas analyzers, mass controller calibrators, chemically selective sensors, and particle detectors are all essential to maintaining the required process cleanliness. Also, environmentally conscious manufacturing will require recycling and reuse of chemicals, not only for waste minimization but also for cost reduction. Chemical generation and reuse will require sensors that can detect impurities at the part per billion level for on-line monitors of chemical purity. These are typified by the chemical sensors discussed in Chapter 6.
Processing of Artificially Structured Semiconductors
A high-priority requirement is the development of noninvasive sensor technologies to sense and control the thickness (down to one atomic layer) of films, especially "superlattice" structures. The use of optical technologies (such as ellipsometry, laser-induced fluorescence, and fiberoptic probes) as an energy transduction medium is rapidly growing in capability and popularity. A material issue relevant to all optical technology is the optoelectronic modulator. It is the interface between optical and electronic components and is the key optoelectronic component for fiberoptic communications. The material currently used in optical modulators is lithium niobate, but its cost is prohibitively high.10 Recently, the French have identified "molecular" optoelectronics research as a potential solution, since molecular materials would reduce the bottle-neck and allow the wide-scale application of optical fibers. Interestingly, one of the processing techniques for growing molecular optoelectronic films is a variation of molecular beam expitaxy—organic molecular beam epitaxy.
AIA (Aerospace Industries Association). 1991. National Advanced Composites Strategic Plan. Washington, D.C.: National Center for Advanced Technologies.
Anderson, R., and C. Grant. 1991. Hercules Aerospace Composite Structures Group, Advanced Fiber Placement of Composite Fuselage Structures. Presentation at the 36th International Symposium, Society for the Advancement of Materials and Processing Engineering (SAMPE), April 1991. Covina, California: SAMPE.
Brennan, T.M., J.Y. Tsao, and B.E. Hammons. 1992. Reactive sticking of As4 during molecular beam homoepitaxy of GaAs, AlAs, and InAs . Journal of Vacuum and Science Technology 10(1):33–45.
Chalmers, S.A., and K.P. Killeen. 1993. Method for accurate growth of vertical-cavity surface-emitting lasers. Applied Physics Letters 62(11):1182–1184.
Chalmers, S.A., L.L. Lear, and K.P. Killeen. 1993. Low resistance wavelength-reproducible p-type (AlGaAs) distributed Bragg reflectors grown by molecular beam epitaxy. Applied Physics Letters 62(14):1585–1587.
Cirino, M., and T. Watson, and D. Hauber. 1991. Composite Structure Fabrication With In-situ Consolidation of APC-2/AS4. Presentation at the 36th International Symposium, SAMPE (Society for the Advancement of Materials and Processing Engineering), April 1991. Covina, California: SAMPE.
Evans, K.R., R. Kaspi, W.T. Cooley, C.R. Jones, and J.S. Solomon. 1993a. Arbitrary Composition Profiles by MBE Using Desorption Mass Spectrometry. San Francisco, California: Material Research Society.
Evans, K.R., R. Kaspi, C.R. Jones, R.E. Sherriff, V. Jogia, and D.C. Reynolds. 1993b. Desorption mass spectrometric control of composition during MBE growth of AlGaAs. Journal of Crystal Growth 127(1/4):523–527.
Gillham, J.K. 1986. Formation and properties of thermosetting and high Tg polymeric materials. Polymer Engineering and Science 26(20):1429–1433.
LeClair, S.R., and F.L. Abrams. 1989. Qualitative process automation. International Journal of Computer Integrated Manufacturing 2(4):205–211.
LeClair, S.R. 1991. Self-directed process control: Can we build material processing systems that know better how to process a material than we do? Concurrent Engineering Journal 1(2):13–22.
Maguire, J.F., P.L. Talley, and J.F. O'Brien. 1992. Laser Fiber Optic Sensor for Polymeric Composites. Final Report No. SWRI Project 06-9572. San Antonio, Texas: Southwest Research Institute.
Maguire, J.F., and P.L. Talley. 1995. Remote raman spectroscopy as a sensor technology in composite materials processing. Journal of Advanced Materials 26(2):1–14.
Miller, A., and R. Van den Nieuwenhuizen. 1993. Full in situ consolidation of graphite fiber, thermoplastic matrix composites by direct electric heating. Pp. H1–H10 in the 10th Thermoplastic Matrix and Low Cost Composites Review. San Diego, California: U.S. Air Force.
Patterson, O.D., S.J. Adams, P.H. Garrett, K.G. Eyink, S.R. LeClair, J.J. Heyob, V. Hunt, K.R. Currie, and T.W. Haas. 1992. Progress Toward a Comprehensive Control System for Molecular Beam Epitaxy. Interim Report No. WL-TR-92-4091. Wright-Patterson Air Force Base, Ohio: Wright Laboratory, Materials Directorate.
SIA (Semiconductor Industry Association). 1994. National Technology Roadmap for Semiconductors. San Jose, California: SIA.
University of Delaware. 1989. Delaware Composites Design Encyclopedia, Vol. 3. University of Delaware Processing and Fabrication Department. Lancaster, Pennsylvania: Technomic Publishing Inc.
Warnock, R., and S.R. LeClair. 1992. Qualitative Process Automation for Autoclave Curing of Composites. Final Report No. WL-TR-92-4085. Wright-Patterson Air Force Base, Ohio: Wright Laboratory, Materials Directorate.
Yolken, H.T., and R. Mehrabian, ed. 1985. A National Forum on the Future of Automated Materials Processing in U.S. Industry: The Role of Sensors. Technical Activities NBSIR 86-3341. Gaithersburg, Maryland: U.S. Department of Commerce, National Bureau of Standards (succeeded by the National Institute of Standards and Technology).
Yolken, H.T., and J.P. Gudas. 1991. Intelligent processing of materials. Technical Activities NISTIR 4693. Gaithersburg, Maryland: U.S. Department of Commerce, National Institute of Standards and Technology.