Publications

Abstract not provided.

We have been studying the use of spectral imagery to locate targets in spectrally interfering backgrounds. In making performance estimates for various sensors it has become evident that some calculations are unreliable because of overfitting. Hence, we began a thorough study of the problem of overfitting in multivariate classification. In this paper we present some model based results describing the problem. From the model we know the ideal covariance matrix, the ideal discriminant vector, and the ideal classification performance. We then investigate how experimental conditions such as noise, number of bands, and number of samples cause discrepancies from the ideal results. We also suggest ways to discover and alleviate overfitting.

Although the theoretical resolution for a conventional optical microscope is about 300 nm, it is normally difficult to obtain satisfactory critical dimension (CD) measurements below about 600 nm. E-beam technology has been popular for sub-500 nm metrology but also has well known limitations. Scanning probe and near-field optical methods have high spatial resolution. Yet they are ill-suited for routine CD metrology of high aspect ratio features because of a combination of short working distances (< 10 nm) and large tips. In this paper the authors present the concept and initial modeling results for a novel near-field optical probe that has the potential of overcoming these limitations. The idea is to observe resonance shifts in a waveguide cavity that arise from the coupling of the evanescent field of the waveguide to perturbations beneath the waveguide plane. The change in resonance frequency is detected as a change in the transmission of a monochromatic probe beam through the waveguide. The transmitted intensity, together with the appropriate signal processing, gives the topography of the perturbation. The model predicts that this probe is capable of determining the width of photoresist lines as small as 100 nm. The working distance is much more practical than other probe techniques at about 100 to 250 nm.

The generation of particles in gas handling systems as a result of corrosion is a major concern in the microelectronics industry. The corrosion can be caused by the presence of trace quantities of water in corrosive gases such as HCl or HBr. FTIR spectroscopy has been shown to be a method that can be made compatible with corrosive gases and is capable of detecting low ppb levels of water vapor. In this report, the application of FTIR spectroscopy combined with classical least squares multivariate calibration to detect trace H{sub 2}O in N{sub 2}, HCl and HBr is discussed. Chapter 2 discusses the gas handling system and instrumentation required to handle corrosive gases. A method of generating a background spectrum useful to the measurements discussed in this report, as well as in other application areas such as gas phase environmental monitoring, is discussed in Chapter 3. Experimental results obtained with the first system are presented in Chapter 4. Those results made it possible to optimize the design options for the construction of a dedicate system for low ppb water vapor determination. These designs options are discussed in Chapter 5. An FTIR prototype accessory was built. In addition, a commercially available evacuable FTIR system was obtained for evaluation. Test results obtained with both systems are discussed in Chapter 6. Experiments dealing with the interaction between H{sub 2}O-HCl and potential improvements to the detection system are discussed in Chapter 7.

Remote sensing by satellite is increasingly important to the national government for treaty verification, battlefield monitoring, and other activities. In addition, civilian oriented applications are increasing in areas such as geology, meteorology, ecology, forestry, and agriculture. Spectral imaging sensors, an important subclass of satellite-borne sensors, have been shown to provide information far superior to that of conventional panchromatic images in many of these applications. However, spectral imaging adds at least two challenges to the already difficult task of viewing the earth from a distance of hundreds of kilometers. First, with numerous spectral channels, the signal-to-noise ratio is decreased in any one channel. Second the data rates of spectral imaging sensors (10 Mbytes/sec, or more) stress the limits of the electronic systems, including the onboard data storage, the downlink bandwidth, and the earthbound image analysis system. This report describes a new concept which the authors have dubbed the information-efficient spectral imaging sensor (ISIS) which addresses these two problems. In addition, it offers the promise of nearly real-time identification of targets.

Water determination in semiconductor process gases is desirable in order to extend the life of gas delivery systems and improve wafer yields. We review our work in applying Fourier transform infrared spectroscopy to this problem, where a 10 ppb detection limit has been demonstrated for water in N2, HCl, and HBr. The potential for optical determination of other contaminants in these gases is discussed. Also, alternative optical spectroscopic approaches are briefly described. Finally, we discuss methods for dealing with interference arising from water in the instrument beam path, yet outside the sample cell.

The feasibility of three different non-destructive and direct methods of evaluating PCB (printed circuit boards) cleanliness was demonstrated. The detection limits associated with each method were established. In addition, the pros and cons of these methods as routine quality control inspection tools were discussed. OSEE (Optically Stimulated Electron Emission) was demonstrated to be a sensitive technique for detection of low levels of flux residues on insulating substances. However, future work including development of rugged OSEE instrumentation will determine whether the PCB industry can accept this technique in a production environment. FTIR (Fourier Transform Infrared) microscopy is a well established technique with well known characteristics. The inability of FTIR to discriminate an organic contaminant from an organic substrate limits its usefulness as a PCB line inspection tool, but it will still remain a technique for the QC/QA laboratory. One advantage of FTIR over the other two techniques described here is its ability to identify the chemical nature of the residue, which is important in Failure Mode Analysis. Optical imaging using sophisticated pattern recognition algorithms was found to be limited to high concentrations of residue. Further work on improved sensor techniques is necessary.

To extend the life of gas delivery systems and improve wafer yields, there is a need for an in-line monitor of H{sub 2}O contamination. Goal of this project is to develop such an instrument, based on infrared spectroscopy, that has a detection limit of 30 ppB or better and costs $50K or less. This year`s work considered the application of Fourier transform infrared (FTIR) spectroscopy to H{sub 2}O detection in N{sub 2} and HCl. Using a modified commercial FTIR spectrometer and a long-path gas cell, a detection limit of about 10 ppB was demonstrated for H{sub 2}O in N{sub 2} and HCl. This includes about a factor of three improvement achieved by applying quantitative multivariate calibration methods to the problem. Absolute calibration of the instrument was established from absorptivities of prominent H{sub 2}O bands between 3600 and 3910 cm{sup {minus}1}. Methods are described to minimize background moisture in the beam path. Spectral region, detector type, resolution, cell type, and path length were optimized. Resolving the narrow H{sub 2}O bands (FWHM {approx} 0.20 cm{sup {minus}1}) is not necessary to achieve optimal sensitivity. In fact, optimal sensitivity is achieved at 2 to 4 cm{sup {minus}1} resolution, allowing the use of an inexpensive interferometer. A much smaller, second generation instrument is described that will have a conservatively estimated detection limit of 1 ppB. Since the present laboratory instrument can be duplicated in its essential parts for about $90K, it is realistic to project a cost of $50K for the new instrument. An accessory for existing FTIR spectrometers was designed that may be marketed for as little as $10K.