Models are highlighted as the key, underpinning element of specification and characterization of silicon wafers in the semiconductor industry. The role of models in interconnecting specifications, metrology, and standards across the domains of processes, equipment, and materials is described. Examples of failing models and new model approaches are provided for lithographic flatness and particle detection.

Measurements of optical scatter are often employed in production line diagnostics for surface roughness of silicon wafers. However, the geometry of the optical scatter instrumentation lacks universal standardization, making it difficult to compare values obtained by instruments made by different manufacturers. The bidirectional reflectance distribution function, on the other hand, is a well-defined quantity, and under conditions usually met with bare silicon wafers, can be related to the power spectral density of the surface roughness. In this paper, we present an approach for characterizing low level optical scatter instrumentation using a spatial frequency response function. Methods for calculating or measuring the response function are presented. Limitations to the validity of the spatial frequency response function are also discussed.

Morphology and topography of mirror polished silicon wafer surfaces were investigated with profiling and light scattering techniques, including in particular Atomic Force Microscope. Roughness <EQ 0.1 nm rms is observed on a lateral scale ranging from about 20 nm up to 10 micrometers in close agreement with values expected from surface topography models composed of atomic steps. Defects consisting of tiny etch pits occur with a low density. These pits are crystal defects delineated by treatment of the silicon wafers with SC1-solution or by polishing. Occurrence of pits in pairs is common but single pits are also observed. The shape of the pits resembles an inverted pyramid with a square or rectangular base. Details of the pits' shape are influenced by the delineation process selected. Pits observed after polishing display a smoother transition region between the very pit itself and the surrounding surface as compared to pits delineated by SC1-treatment. Smoother pits have a smaller effective cross section for light scattering with respect to specific types of surface inspection instruments. This results in significant differences in counts of light scattering defects when surface inspection tools of various suppliers are used. The effect can be explained by considering the spatial frequency bandwidth of the various surface inspection instruments used. Bandwidth differences are also responsible for differences in rms roughness values as reported by profiling instruments.

Especially for wafers, hard disks and flat panel displays fast and accurate technical means for roughness characterization are needed. However, speed and accuracy are contradictory. Generally speaking, fast roughness sensors are not accurate, and precise instruments are slow. It turned out in the last years that with multi aperture fiber optic sensors which acquire ARS/TIS data a very fast estimation of surface roughness is possible. But it is rather difficult to convince e.g. chip manufacturers that the results of such sensors are reliable, because there are no accepted international standards for these kinds of optical measurements. Therefore we decided to establish a setup of our ARS/TIS sensor for roughness characterization and an instrument for roughness measurement in a cleanroom consisting of the following parts: (1) 200 X 200 mm stages, speed 0.4 ms^-1, +/- 1 micron accuracy, acceleration 1 g; (2) visual inspection head consisting of 50 X objective and CCD camera; (3) AFM scan head; (4) ARS/TIS fiber optic sensor; and (5) laminar box. Topics of the paper are measurement philosophy, specs of the setup, architecture of the fiber optic ARS/TIS head, as well as data processing algorithms and software.

An approach to increase the signal-to-noise ratio of the scatter data (hence its accuracy) when studying super-smooth surfaces is proposed. It consists of increasing the power of the illuminating source. Since stray scattering is no longer negligible in such an operation, a solution to alter the instrument signature is given. To prove their stability, the cross section data from a silicon carbide mirror are used in a singular manner to extract roughness parameters mainly the root mean square roughness and the correlation length. Results are satisfactory as they feature low dispersion and are consistent with the approximation used.

A scheme based on the distorted wave Born approximation has been developed to model the off-specular x-ray reflectivity from flat surfaces with compositional and topographic fluctuations. To verify this theoretical work, silicon wafers coated with evenly spaced aluminum lines were chosen as the test samples. Good agreement is found between the calculationed and the experimental results. In addition, a gross difference in the off-specular spectra was observed from two test samples different only in their surface roughness; this observation demonstrates the potential of using off-specular x-ray reflectivity for quality control measurements.

This paper derives the Mueller scattering matrices for two topographic scattering models--the Rayleigh-Rice or perturbation model and the geometrical-optics or facet model. The results are used to predict the polarimetric properties of the `haze' on silicon-wafer surfaces.

A number of physical standards are available for the measurement of surface roughness, but none of them are applicable to use with the very smooth surfaces being manufactured in the semiconductor, computer disk and flat panel display industries. This paper reviews some of the issues and suggests possible approaches for realizing such standards. Suggestions include grating like surfaces, very square steps and isotropic (polished) silicon.

Microroughness, or haze, on wafer surfaces can mask the detection of particles by scanning surface inspection systems (SSIS). The ability of silicon dioxide or other films to function efficiently as insulators depends partially on the underlying microroughness of the silicon surface. For thin oxides, breakdown voltages are reduced commensurately with increased levels of microroughness. There are similar effects on film layers deposited in later processing steps, and an effect on bonding for silicon-on- insulator applications. The disk drive industry depends on a steady transfer rate of data from the recording medium. Imparted surface texture must be carefully controlled since it is in conflict with the desire to have the head in close proximity to the recording surface; however, too fine of a polish can lead to stick-slip or blocking. Additionally, the surface texture must be very uniform across the face of the recording media. The flat panel display industry, with their ubiquitous screens now so common in laptop computers, relies on an orientation layer with a precise amount of microroughness on the substrates. In these and numerous other applications, a well characterized surface is paramount to high production yields. And yet, the most often used of these measurements, rms microroughness, is grossly misunderstood. Surface roughness is not a unique number nor is it an intrinsic surface property. Roughness measurements depend on the parameters of the instrument used for measurement--whether that be an optical or mechanical profiler, a SSIS used for haze detection, or an atomic force microscope. Each of these instruments may give very different values from exactly the same surface. A novel approach to developing a practical haze standard has been employed by photolithographically etching features to as little as 1 nm deep into the surface of 150 mm silicon wafers. To prevent a wafer scanner from detecting these features as particles or other light scattering events, a high surface density (4 X 10⁶ features/cm²) is produced such that the distance between features is much less than the spatial resolution of current instruments (typically 50 micrometers - 100 micrometers ). Once the wafers have been fabricated, the haze or microroughness level detected by a given scanning instrument may then be calculated from the Power Spectral Density function plots generated for each wafer by various techniques (such as angle resolved light scattering or atomic force microscopy). The amount of simulated haze produced by this method is a function of the depth of etch into the silicon surface.

We describe a computer aided surface classification method with the scanning confocal microscope for observing a high aspect ratio device with 3D structure. As such device, we make use of Digital Mirror Device (DMD). To eliminate the distortion in the display image with DMD, the micro-mirror of DMD should be processed so that they form an exactly flat plane, and the condition of mirrors should be monitored after manufacturing DMD. In monitoring with the conventional optical microscope, it is hard to observe the accurate 3D condition due to very limited depth of focus and poor resolution. SEM can not be used to confirm the high speed motion of micro-mirror by the external digital driving signal to generate the display image, because SEM needs the closed vacuums environment. Solving these issues, we have developed the 3D visualizing technology with the scanning confocal microscope. The scanning confocal microscope can achieve the resolution superior to the conventional optical microscope. Furthermore we can check the tilted micro-mirror of DMD in the atmosphere. Considering that a raw image of the scanning confocal microscope is of higher resolution, we treated a raw image as an optical tomographic image, and developed a 3D visualizing method by accumulating such thin optical tomographic images of the scanning confocal microscope.

In response to the semiconductor industry's need for both smaller calibration particles and more accurately sized larger particles, a joint SEMATECH, National Institute of Standards and Technology, and VLSI Standards, Inc. project was initiated to accurately characterize 10 monodisperse polystyrene sphere suspensions covering the particle diameter ranges from 70 to 900 nm. The sizing analysis is being performed by electrical mobility analysis with a modified flow system to enable the measurement of the width of narrow size distributions. Results are presented on the mean size and the width of the distribution for candidate samples provided by five suppliers. The target sizes for the first set of particles are 72 nm, 87 nm, 125 nm, and 180 nm. Challenges for detecting `real world' particles are discussed including quantitative examples of the effects of refractive index, layered structure, and non-spherical shape on the light scattered by a particle.

Even though automated particle inspection sensitivities exceed 0.100 micrometers , visual inspections continue to flourish in the silicon wafer manufacturing facilities. These inspections continue due to the inability of Scanning Surface Inspection Systems (SSIS) to detect and identify a variety of wafer defects. Single integrating collection schemes have not provided sufficient information to accomplish this task. To realize full automated wafer inspection, additional inspection data must be obtained by the inspection process. Visual inspection has made great use of the near specular collection. SSISs must make use of this region. The employment of near specular inspection will greatly enhance the detection capabilities of these tools. This information must then be processed in a way to allow defect identification and sorting.

While angle-resolved scatter has been extensively investigated for particles on smooth and rough semiconductor surfaces by Bawolek and Warner, similar fundamental studies for particles on patterned surfaces are quite limited. In this paper we report results of research on the effects of adjacent surface features (simulating IC patterns) on scattering by particles. Experimental measurements of angle- resolved scattering signatures of individual submicron particles on a test wafer are presented. In particular, the effects of the relative position of the particle with respect to the pattern are shown. The results provide some fundamental insight into the potential particle sizing errors associated with particle detection on patterned surfaces.

Detection of particle contaminants on patterned wafers is important for increasing yield during the silicon wafer manufacturing process. Surface scanning inspection systems are used to detect contamination by measuring scattering from coherent light incident on the wafer surfaces. To aid in the design of these inspection systems, a code based on the coupled-dipole method has been developed to predict scattering from features on surfaces. To validate the code, we show comparisons between experimental results and numerical predictions of scattering characteristics of patterned structures found on the Arizona State University/Semiconductor Research Corporation block of the SEMATECH patterned wafer defect standard die developed by VLSI Standards, Inc. The patterned structures considered here are SiO₂ line features and cornered features on silicon substrates. Particle contamination is emulated with deposition of PSL spheres near these structures. Comparisons of the differential scattering cross-section are made between experiment and the computational results. Close agreement within a factor of 2 to 3 is found for the cases considered.

Measurement equipment for geometry, shape, surface finish of bare silicon wafers are highly sensitive tools operating partly at their capability limits. This results in performance problems which have to be solved if such equipment is to be of benefit for developing future generations of silicon wafers with specifications according to the SIA roadmap for semiconductors. The instruments for measuring the mentioned parameter belong to a class of tools the output of which is strongly influenced by their bandwidth. The requirements for future measurement equipment for specific measurement tasks as well as for generic capabilities which have to be met by all tools operating in a production ambient are outlined.

An important parameter in the production of flat panel displays is the waviness, or micro-corrugation, of the substrate surface. It describes surface deviations in the mid spatial frequency range between roughness and global shape. Typically, the waviness has to be determined to an accuracy in the 5 nm-range. A technique is presented to measure the waviness optically in a non-contact fashion along a profile 14' long. Thus large panel substrates can be measured. Special issues to consider are: (1) suppression of the light reflected from the back surface of the panel substrate, (2) large required dynamic range due to the overall shape of the panel of more than 100 micrometers , (3) the flexibility of the panel substrates requiring well designed fixturing to avoid bending and pick-up of vibrations. The described technique is based on an extension of a phase- shifting white-light Mach-Zehnder interferometer which has been used successfully in the characterization of thin glass disk substrates.

A new technique is proposed to provide a profile measurement, especially for computer discs, wafers or glass substrates that are highly accurate in flatness. A reflection moire technique using phase shifting method is tried to this flatness measurement. To apply this method to large size samples such as LCD flat panel displays, two trials extension of measurement size and elimination of the reflected light from the back plane of the glass substrate, are developed. One is to combine small areas one after another the flatness of which is measured in advance. The other is using UV wavelength. A wavelength of 313 nm, which is absorbed into the LCD glass substrate, adopted as the light source, and synthetic fused silica lenses are used in optics. It is possible to measure the flatness of the front plane. The technique, which produces much higher accuracy than conventional techniques, can be available in various of optical measurements.

The evanescent field coupling characteristics between a single-mode half coupler and a thin-film planar waveguide (PWG) placed on the substrate is discussed. The throughput fiber power T changes with the variation of the effetive thickness of the film. Experiments for different thicknesses of the PWGs are carried out. The . thickness of the spacer deposited on the substrate or the roughness of the substrate with fixed spacer, controls the effective thickness of the PWG. Thus, T measures the roughness quality of the substrate and also the thickness of the spacer. The substrate was fused silica for the standard telecommunication fiber as half coupler. However, coupling fiber was also used in the case of germania-silica substrate for matching the propagation constants fo the fiber and the substrate. The variation of T with the change of refractive index of PWG and also the substrate are also shown. Theoretical analysis was carried out and a good agreement is observed with the experimental results.

In this paper, an instrument for measuring flatness derivation by using an aligned laser beam and CCD detector is described. A stable aligned laser beam is used to form a datum plane in the instrument by rotating shaft and a pentagonal prism. Using a linear array CCD detector as the sensing probe, the system can detect the deviation of height related to the datum plane directly and absolutely. The experiment result shows that the measurement accuracy ofthe instrument is 2.3x r6. Keyword: Flatness, Laser beam, CCD detector