Intelligent Processing, Sensors, & Controls:
From the time of its inception ITN had recognized that today's competitive global economy requires continued advances in manufacturing technology to achieve and maintain cost advantages. Critical to meeting this challenge is to incorporate appropriate control and sensor technologies to ensure best product quality and maximum yield at each stage of production.

ITN has proven its strengths in the areas of

• controls,
• process modeling,
• fault management,
• software development and integration, and
• development and integration of unique sensors 

These strengths are leading to a wide range of products, such as integrated real-time control systems, suite of sensors, fault-tolerance, condition monitoring of equipment, software utilities/tools, and consulting services. ITN's efforts can help any production facility to remain globally competitive by maximizing yield, improving product quality, reducing downtimes, and optimizing processes, including the semiconductor industry. Some of the concepts developed can also be applied to a wide variety of applications.

Control:
ITN is actively involved in the intelligent processing of materials that encompasses methodologies of controlling the processing and manufacture of materials. As shown in figures 1 and 2, intelligent processing involves in-situ sensing and control of microstructure/properties of the material and process variables, and process optimization. To ensure that superior thin film products are produced at high yields it is necessary to control the process at each stage of the production. To enable this control, a wide range of sensors is used, which must communicate with the control hardware/software. For example, the deposition systems currently used by ITN and its sister company, Global Solar Energy (GSE) use simple thermocouples and fluid flow meters as well as more complex instruments such as ion pressure gauges, atomic absorption spectroscopy, X-Ray fluorescence spectrometers, and parallel detection spectroscopic ellipsometers. Using these in-house developed sensors and dynamic models of the processes ITN has successfully designed and implemented thickness controllers for each element in the complex thin-film copper-indium-gallium-diselenide (CIGS) photovoltaics systems. Figure 3 shows a representative plot of the thickness.

Figure 1 – Framework of Intelligent Processing of Materials

Figure 2 – Architecture of a Controller

Sensors
Over the past several years, ITN in conjunction with experts at The University of Delaware (IEC), National Renewable Energy Laboratory, University of Virginia, and others have investigated a number of sensors for use as in-situ real time process control monitors for manufacturing thin films (e.g. Photovoltaics, giant-magneto resistance films, multiple thin film stacks, and flat-panel displays). The goal was to identify measurement techniques that had the greatest commercial availability and/or those that could provide in-situ (both inside and outside the deposition zone), non-destructive evaluation and process control for rigid and flexible thin film manufacturing. The technique(s) had to be adaptable to the design and manufacture of an inexpensive, versatile, robust, on-line sensor that could provide in-situ process control information in the harsh environment of a vapor deposition chamber where heat and coating of sensor components can damage or interfere with the instrument.

Table 1: Summary of Sensors/Techniques Investigated for CIGS Process Control

In general ITN has detailed knowledge of most of the standard ex-situ characterization techniques. Furthermore, ITN is also very knowledgeable in most of the potential in-situ characterization techniques that can be implemented for real-time process control. Table 1 highlights the advantages and disadvantages of the potential in-situ techniques that ITN has or is presently incorporating for process control. Typically, ITN will purchase commercially available sensors, when available, and adapt them for operation in specific systems. However, when a sensor is not available, ITN has also developed sensors for specific and general applications; including: spectroscopic ellipsometry and x-ray florescence.

ITN has successfully developed two unique sensors of its own, the X-Ray Fluorescence (XRF) (link) and Parallel Detection Spectroscopic Ellipsometer (PDSE) (link) for the thin-film deposition systems as described in the following sections.

X-Ray Fluorescence (XRF)
X-ray fluorescence measurements are performed by illuminating a portion of the sample with x-rays and then measuring the energy and count rate of the fluoresced x-rays. Incident x-ray photons cause electrons to be ejected from atoms in the sample. As the remaining electrons fill the newly created vacancies, excess energy from relaxing electrons is emitted as x-rays. The energy of these fluoresced x-rays corresponds to the energy change of the electron transition, and therefore each element fluoresces at a characteristic set of energies. X-rays resulting from the most probable transitions terminating in the K shell are known as "Ka" x-rays, and x-rays resulting from the most probable transitions terminating in the L shell are known as "La" x-rays. Fluorescence occurring due to direct excitation by x-rays from the x-ray source is termed "primary fluorescence". Fluorescence from an atom excited by x-rays fluoresced from the other elements in the sample is termed "secondary fluorescence". For in-situ monitoring, fluoresced x-ray energies and rates are measured with a solid-state energy-dispersive detector, due to its speed and compact size. Figure 1 illustrates an XRF instrument developed by ITN and successfully used in production systems while Figure 2 shows its schematic.

Figure 1: XRF Instrument Developed by ITN Used on a System

Typical XRF systems are installed as an accessory on scanning electron microscopes, or as self-contained desktop and portable units for soils and metals analysis. A few important differences exist between the requirements for XRF systems in these typical applications and XRF systems for to be used for in-situ composition monitoring. First, during in-situ monitoring of CIGS growth, measurements are made on large samples containing substantial amounts of known elements, whereas typical XRF systems are designed to be able to measure small samples possibly containing trace amounts of unknown elements. Further restrictions are imposed by CIGS deposition chamber geometry, the presence of Se vapor, elevated temperatures, and measurement time requirements during in-situ monitoring.

Figure 2: Schematic of the XRF Developed at ITN

Thus, XRF in itself is not a new measurement. ITN's design improves upon the basic XRF through a number of unique features, including:

• Hardware include protection of the sensor from the deposition environment,
• Use of a sensor-to-sample distance appropriate to deposition chambers,
• Use of only low-cost components operating at room temperature,
• Analysis include one-sample calibration that gives valid results over a wide range of compositions,
• Real-time CIGS analysis,
• Compensation for variations in substrate location and x-ray tube current drift by using the substrate signal, and
• Use of XRF as a sensor for real-time closed-loop control of deposition.
• The application of XRF to CIGS deposition allows the use of innovative hardware and analysis because the elements present in the film are known prior to measurement, 1% precision is sufficient, and recent advances in x-ray tube and detector manufacture have occurred.

Parallel Detection Spectroscopic Ellipsometer (PDSE)
For thin film manufacturing that is sensitive to processing conditions, sensors are needed to measure film stoichiometry, and film optical/electronic properties. Measurements must be non-invasive, non-perturbing, and remote (non-contacting) from the vapor and film surface. For complex thin-films, where optical properties, and film morphology and thickness are very sensitive to processing conditions, simpler optical measurement techniques like reflectometry and interferometry cannot uniquely determine all the optical and physical film properties, simultaneously, even with spectroscopic techniques. Only by monitoring the polarization state of the light and how this is modified through the interaction with the sample (ellipsometry) to determine both the amplitude and phase, is it possible to uniquely determine the refractive index and structural features of a film, simultaneously. For this reason and others, polarization state analysis has become one of the most valuable methods in evaluating film performance of products that depends critically on the electron band structure. Furthermore, since the polarization state of light does not change as it passes through thin amorphous or non-birefringent films at normal incidence, a technique like spectroscopic ellipsometry will not be affected by undesired coating on optical windows.

Figure 1: PDSE Hardware and Calibration Data Comparison to a Conventional Ellipsometer. The data indicate that the PDSE can obtain the same information in 3 ms that a conventional ellipsometer took several minutes to collect with as good or better accuracy.

ITN has developed a PDSE, shown in Figure 1, to provide the film property measurements needed for advanced in-situ process control. Operational PDSE instruments that use four spectrometers to detect light between 250 and ~1700 nm have been adapted to provide in-situ real-time thin film property measurements from rigid and moving flexible samples. The significant advantages and film property information that the PDSE is capable of providing include:

ITN's PDSE sensor provides more than an order of magnitude faster data acquisition times than conventional spectroscopic ellipsometers enabling real-time analysis of films as they are being deposited. The compact detection heads with no moving parts and flexible design enables PDSE systems to be incorporated directly into vacuum deposition chambers with little to no modifications. No other spectroscopic ellipsometry system is presently vacuum compatible. Since there is no vibration noise from moving parts and the four polarization signals are simultaneous detected, systematic and random errors due to signal drift during serial data acquisition are eliminated. Thus, the PDSE sensor is intrinsically more accurate. The elimination of moving parts and a simple design significantly reduce the instrument costs. Furthermore, the unique hardware design that allows sensor placement inside the deposition system enables all levels of in-situ real-time data interpretation including minimal data analysis approaches to be used. By measuring changes in the polarization state of reflecting light, the PDSE sensor provides cost-effective in-line sensing for intelligent process control and can detect critical product variables that directly relate to film performance.

Thus incorporation of PDSE will increase the overall quality of the materials, decrease failure rates, and increase product yield and performance. These measures will greatly decrease costs by minimizing deposition of expensive materials, excess handling for characterization and quality control, and ultimately environmental damage from wasted materials. The decreased cost and increased performance that can be gained from implementing the PDSE sensor for process control can ultimately help raise the relatively poor performance associated with some state-of-the-art materials, enabling additional commercialization of these products.

In general, the PDSE has been developed as an in-line sensor that will provide real-time control for commercial thin-film deposition processing. Some of the applications include: ·

• Thin-Film photovoltaics (original application),
• High temperature superconductors,
• Flat panel displays,
• Layered semiconductors 
  • Doping, annealing, deposition
• Coatings
  •  Optical, molecular, oxides

Application of PDSE for the giant magnetoresistive (GMR) provides additional insight into its capabilities. For GMR, the PDSE provides the necessary information so that feedback/feedforward control strategies can be devised and implemented to increase the overall quality of GMR multi-layer stack materials, decrease failure/low GMR rates, and increase product yield and performance.

a.	b.

Figure 2: Predictions about the Sensitivity of the PDSE to Cu Layer Thickness at (a) 450 nm and (b) from 250 to 800 nm. (c) and (d) are results for a GMR stack that show the amount of change in ? and ? for a 1 angstrom change in Cu thickness. The data indicate that if difference measurements can be made during film growth, then the PDSE will have sufficient accuracy to provide process control information.

Initial spectroscopic ellipsometry measurements of GMR stack and individual material layers, shown in Figure 2, indicates that the ellipsometry parameter, Psi, has a near linear relationship with the Cu layer thickness and can determine thickness to a monolayer or less if Psi can be accurately determined to within 0.3 degrees. Furthermore, these initial evaluations indicated that the PDSE is extremely sensitive to the specific optical properties of the individual layers, suggesting that small changes in the optical properties due to interfacial mixing or change in the physical structure should be observable.

While the SE data of the GMR stack indicates that the PDSE will be able to obtain information about all the layers, pragmatically it will be very difficult to separately determine significant changes in more than one layer. Thus the true power of the PDSE is its ability to provide real-time in-situ information about each layer as it is being deposited. Detailed analysis of each layer deposition will provide a maximum amount of information about each layer and its interaction with the previously deposited layers. True real-time process control strategies can then be based on "minimal data approach" that relies on measurement difference comparisons between each processing step, as illustrated in Figure 2 c & d). The initial SE data analysis performed here indicates that the PDSE will have sufficient accuracy and sensitivity to provide the information required for this "minimal data" process control strategy for GMR deposition.

While each optical/electronic process has its own intrinsic issues that must be addressed, the initial evaluation of the GMR samples indicates that the PDSE will be very useful for providing the information needed to implement a "minimal data" approach for many different multi-layer optical/electronic coatings. In fact, the PDSE may be the only sensor that can provide the in-situ real-time information about each film layer as it is being deposited, thus providing the only real information, in-situ or ex-situ, that can be used to evaluate device performance and tie that performance directly to process parameters.

Fault-Tolerance and Fault-Tolerant Control Systems
A typical process control system results in unsatisfactory performance or even instability in the event of malfunctions in sensors, actuators, or other components of the system. This problem is very critical when the sensors and other components are exposed to harsh environments. In order to overcome these limitations ITN has recently secured government funding for a project that includes developing controllers that are capable of tolerating system malfunctions and maintains desirable performance and stability properties.

Modeling
ITN has extensive experience in the area of modeling and model reduction. For example, for the CIGS deposition system it was important to know the surface temperature of the different melt in the effusion source, but due to the harsh environments it was not practical to monitor the temperature directly. Instead thermocouples are mounted in the walls of the effusion sources. Sophisticated thermal models of the sources were developed to predict the surface (pool) temperatures based on the thermocouple and power supplied, all used in real-time. As shown in Figure 1 the model predicts the thermocouple reading very well. It also predicts the pool temperature, which, in turn, is fed to other models in the hierarchy of models for the purpose of process control. These models consider all the relevant dynamics, such as depletion of material with time.

Figure 1: Comparison of Simulation and Actual Temperatures

Condition Monitoring of Equipment
It is not uncommon to see that an equipment has at least 5% unplanned downtime due to unforeseen failures. This could easily represent up to 3% of revenue. ITN is trying to make intelligent use of information generated by different sensors in a system to provide early signs of component failures resulting in elimination of unplanned downtimes and allow planned preventive maintenance.