A common cause of nonlinearity in lithographic metrology with SEMs is charge accumulation on photoresist structures surrounding the features to be measured. This phenomenon has been observed to produce strikingly different results on three low-voltage (1 kV) SEMs evaluated under different operating conditions. Features examined were isolated lines, lines in gratings, isolated spaces, and contact holes which ranged from 0.5-1.3 micrometers in 0.1 micrometers increments. Critical dimension measurements at the base of photoresist structures were obtained from image linescans using algorithms indigenous to the systems. The linescans exhibited various degrees of intensity blooming, scan asymmetry, and image inversion as a function of operating conditions.

A Monte-Carlo simulation program has been modified to allow specifying the specimen surface as a series of regions with arbitrary shape and composition. It also models the production of the secondary electron signal. This has been applied to a systematic series of experiments with 0.5 and 1.0 micrometers lines of photoresist, and compared to experimental measurements using an SEM. The qualitative agreement indicates that the model can be used to study the effects of variations in operating conditions such as accelerating voltage, as well as the effect of changes in specimen geometry or composition.

A new figure of merit, the critical dimension capability factor, of CDC, is described. The CDC incorporates measurements taken over a range of linewidths and over a range of process variations which simulate normal and extreme process operating conditions. Under these conditions CDC uniquely quantifies the capability of the measurement instrument on a given substrate and for a given set of parameter settings. CDC is calculated by performing a linear regression between measurements generated by the instrument under test (IUT) and a set of reference values (internally generated standard values). The mean square error (MSE) between the regression line and the observed values is then partitioned into components which estimate the contribution to the MSE from various sources based on a rigorous statistical analysis. The final CDC value is defined as the linewidth to uncertainty ratio and is a function of uncertainty introduced in the characterization procedure as well as the uncertainty introduced when the IUT makes a measurement in practice. Since the CDC is a function of the overall uncertainty in the measurements of the IUT relative to the reference values, it can legitimately be compared from one instrument to another and used to evaluate alternative measurement methods and technologies.

Masks used for the manufacture of integrated circuits by x-ray lithography can be calibrated and inspected in a scanning electron microscope by using the transmitted electron detection mode. By their nature, these masks present a measurement subject unique from most (if not all) other objects in semiconductor processing because the support membrane is, by design, x-ray transparent. This characteristic can be used as an advantage in electron beam- based mask metrology since, depending upon the incident electron beam voltages, substrate composition and substrate thickness, the membrane can also be essentially electron transparent. The areas of the mask where the absorber structures are located are essentially x-ray opaque as well as electron opaque. Viewing the sample from a perspective below an x-ray mask can provide excellent electron signal contrast (depending upon the instrument conditions) between the absorber structure and the membrane. Thus, the mask can be viewed in the transmitted electron detection mode of the scanning electron microscope, and precise and potentially accurate dimensional measurements can be made. One unique advantage to this is that in the transmitted electron detection mode, the modeling of the electron beam\specimen interaction becomes far less difficult than in the modeling of typical secondary electron images of opaque objects. The inelastically scattered beam electrons and the low- energy secondary electrons can be excluded from the detector and, therefore, need not be accurately modeled. Therefore, absorber structure width (linewidth) measurement standards can be potentially calibrated with less difficulty and higher accuracy than standards calibrated by more conventional means. The transmitted electron detection mode is also useful, because of the high contrast of the image, for the determination of mask defects and high- density particle detection as well as for registration measurements.

Linewidth measurements of clear lines on anti-reflection coated photomasks are obtained with a photometric optical microscope and a scanning electron microscope (SEM). Both microscopes have stages fitted with helium-neon laser interferometers and thus provide directly traceable length measurements. In both instruments, measurements of pitches are independent of the choice of image edge intensity threshold, while linewidth measurements are greatly affected by the threshold levels. Simple scalar optical theory predicts that in the absence of aberrations, the position of the edge of a totally opaque film should be at the 25% intensity threshold for coherent illumination. The 50% backscattered electron intensity threshold has proved to be a repeatable point for measurements as it is very insensitive to such parameters as beam diameter, accelerating voltage and exact focus. The relationship between the 50% threshold and the dimensions of the sample itself is obtained with the aid of Monte Carlo electron trajectory calculations. Linewidth measurements made with the two different techniques on the same photomask are presented with particular interest in lines with widths below the resolving power of the optical microscope operated at visible wavelengths and using coherent illumination. The measurement agreement is better than 0.03 micrometers for lines which are fully resolved and increasingly worse for lines with widths below 0.7 micrometers . The degree to which theory predicts agreement between the two techniques and the resulting confidence in the SEM measurements are discussed.

Under low dose conditions, low-voltage SEMs can measure the bottom width of a resist profile accurately over a wide range of resist side wall angles, thus allowing a accurate, precise and fast measurement of focus-exposure matrices. It is also shown that the measured data fits a fixed parameter model well. By means of statistically leveraged experimental design techniques, a robust focus-exposure-response surface may be generated with as few as nine measurements. This approach greatly reduces measurement time, resulting in high-speed stepper setup. Results are shown for different experimental designs and different low- and-high voltage SEMs.

Progress is reported on development of a joint European electron beam based CD metrology instrument for sub-0.5-micrometers lines. The current work is funded as a project within the European Strategic Programme for Research and Development in Information Technology (ESPRIT), a department within the European Economic Community (EEC) and is being undertaken by a consortium of two equipment manufacturers, three IC manufacturers or users, and four research institutes. The project has been running for a period of 15 months, and this time has been used to undertake R&D of the most critical and high risk elements. The following major elements are covered: (1) Defining the user needs, including the operating environment. This has resulted in the specification of an 8' cassette-cassette, 20 wph instrument capable of both measurement and inspection in a hostile environment vis-a-vis stray fields, acoustic and floor vibrations. (2) Using advanced Monte-Carlo techniques and finite field analysis, modelling of the total image forming chain for low energy electrons. (3) Development of new measurement algorithms based on the information from (2). (4) Development of various hardware elements. The task of integrating the above elements into a full prototype including a wafer handling system control and user qualification will take place in the second half of the project which has received additional funding support from the Joint European Submicron Silicon Initiative (JESSI).

SEM cross-sectional (CS) imaging widely practiced in CD measurements of sub- micron (down to .10 micrometers ) features will be reviewed with emphasis on the accuracy with which these measurements represent the actual dimensions. Since the CS imaging mode is commonly employed to establish a correct CD linescan measuring algorithm, the magnitude of systematic error associated with CS imaging at specific SEM settings will be addressed. The image formation mechanism developed by K.-R. Peters was invoked in the investigation of the electron-scattering signal generated by topographic contrast in CS imaging. The main result of this study--the systematic error in the CS imaging mode, at the same experimental conditions, is larger than that in the normal incidence linescan mode--has been found to markedly contradict the commonly accepted metro logical assumption applied to CS CD measurements. Theoretical and experimental results supporting this conclusion will be presented and discussed.

The purpose of this work was to extend the design criteria of electrical test structures to the half-micron linewidth region. At 0.5 micrometers , process limitations place constraints on the functionality and usefulness of electrical test structures based on conventional design criteria. In particular, small total variations from lens aberrations/distortions and proximity/corner rounding effects in the patterning of the smallest lines achievable (less than 0.5 micrometers ) can result in failure of the structure. This was particularly significant when orthogonal voltage taps at minimum design geometries were used. As geometries decrease in size and control over the process and equipment tightens, the intrinsic error in conventional structures increases as a percentage of the total measurement. The design criteria of these structures have been further modified and improved in order to address known lithographic limitations and establish a more process tolerant design. The resulting measurement precision accommodating these changes is discussed to provide the framework for achieving the highest practical performance attainable from both the test structure and the measurement system.

A comparison of SEM measurements vs. electrical measurements of contact holes is presented. In-line SEM measurements on a Hitachi S6000 and measurements of micrographs from an off-line JEOL 845 SEM are compared to electrical measurements on a Prometrix LithoMapR system for metrology of contacts down to 0.25 micrometers size. The electrical measurements of contacts through focus/exposure variations on the stepper are shown to correlate very well with SEM measurements. Electrical measurement of contact holes in conductive films is shown to reflect actual process latitudes on oxide wafers, allowing electrical metrology to be used in optimizing lithography processes.

This investigation studies linewidth metrology techniques and compares SEM measurement results with an electrical linewidth probe procedure. Calibration offsets between reticle, resist, etched, and electrical probe dimensions are compared for the 500-nm nominal isolated and grouped images. These individual probe-site measurements, as monitored with a low-voltage SEM through this process, are then compared to individual-site and full-field electrical data.

A new tool for optical metrology, the Mirau Correlation Microscope (MCM), it introduced. The basic principle of this device is to employ an interference microscope with a temporally and spatially incoherent illumination source, and to use as the detected output the interference signal between the beams reflected from the object and from a reference mirror, respectively. Phase images and cross-sectional images of integrated circuits are obtained. Critical dimension measurements on photoresist linewidths are demonstrated and features with sizes down to 0.4 micrometers have been measured consistently.

Two approaches to pattern recognition of the width and slope of trenches in photoresist and silicon are described. One method depends on the comparison of the experimental linescans with theoretical calculations of the linescans obtained with either a real-time confocal scanning optical microscope (RSOM) or the Mirau Correlation Microscope (MCM). Comparison of theory and experiments for trenches in photoresist and silicon are in excellent agreement for all trench widths from 0.5 micrometers to 2 micrometers . A second method depends on the use of correlation techniques using the MCM and has enables us to obtain good linearity in trench width measurements down to a width of 0.35 micrometers in 1 micrometers thick photoresist.

Linearity of response is one of the most important features of a measurement system. Linearity implies that accurate linewidths can be obtained from measured values knowing only the slope and offset of the data with respect to reference data taken with another, presumably more accurate, instrument. A first-order linear regression of the data yields the slope, offset, and estimate of the goodness of fit. Ideally, the slope is near unity, so that the magnification scales of the two instruments agree. The offset is considerably less important since, in IC process control, absolute changes in linewidth are often of more concern than the linewidths themselves. This paper demonstrates by simulation and experiment that the linearity (R-Squared) of an optical microscope depends not only upon the characteristics of the tool but also upon the characteristics of the object being measured. In virtually all optical microscopes, transparent structures support waveguide resonant eigenmodes which are strongly affected by geometry and contribute substantially to non- linearities in response. For isolated lines, nonlinearities are found to occur especially at certain widths where the eigenfunctions change rapidly with small change in width. The theory of these singular points is presented. The authors demonstrate that the coherence microscope, which uses both phase and amplitude information, has a potential advantage over brightfield and confocal microscopes in dealing with these problems. The introduction of a 'complex phase filter' in the measurement algorithm greatly reduces unwanted phase noise and its concomitant contribution to non-linearity. The ability to simulate the optical images and resulting measurement non-linearities offers promise in improving understanding and accuracy of optical metrology.

A massively parallel computer simulation algorithm is used to investigate electromagnetic scattering and optical imaging issues related to linewidth measurement of polysilicon gate structures. The algorithm simulates scattering of normally incident, TE polarized illumination by inhomogeneous, nonplanar, 2-dimensional topography. The thickness and extinction coefficient of the polysilicon layer (nominally 450 nm thick) are determined from reflectivity versus wavelength data gathered using a Nanometrics Nanospec/DUV microspectrophotometer. These parameters are used as part of the input for the simulator which computes the diffraction efficiencies of the polysilicon gate structure. A spectral component-weighting technique was applied for estimating optical microscope images based on the diffraction efficiencies from normally incident illumination. Simulated image profiles of isolated edges and 1.2 micrometers lines are then compared with images obtained using an NBS-type optical microscope at VLSI Standards, Inc. and show good agreement. The dramatic effects of an 8 nm variation in polysilicon thickness on the electric field distribution within the gate, on reflectivity and on the image profile are illustrated. Also, the effect of focus position is shown by comparing measured and simulated image profiles at different focus offsets.

IC fabrication problems grow as nominal feature sizes shrink, due in large part to fundamental optical diffraction limits. Currently, one of the most pressing needs is robust critical dimension measurement. However, optical methods must be refined for this scale of submicron metrology, particularly in the case of thick features. This paper examines the problem of reflected light microscopy for nominal 1 micron high lines on silicon using 2-D, time-domain finite element simulations. The experimental basis is a prototype line width standard that is characterized using optical, contact, and SEM measurements. Microscope and simulated images are compared for 1 and 3 micron wide lines. Good first-order correlation is found between real and synthetic images but model uncertainties need to be reduced and microscope aberrations need to be quantified before second-order differences can be eliminated. Numerical experiments are used to relate images to resonance patterns in the feature; determine the strength of evanescent waves near the line; and contrast isolated and periodic line images as a function of pitch.

A novel statistically designed tracking technique is described which utilizes multiple exposure systems and photoresists to decompose tool, material, and process contributions to variation in lithography. The application of this technique is described as a means of characterizing an overall lithography process. In one application, the primary source of variation in the photo process was attributed to resists and resist processes rather than related to exposure tool. Process variation in one resist system was tracked unambiguously to changes in developer normality. Finally, results from this work extended previous studies showing that open frame exposure measurements are sensitive measures of photoresist process variation for both conventional and chemically amplified resists.

Fiber-optic-based reflectivity measurement during spray or spray/puddle development is effective for accurate critical dimension (CD) control. However, there are some difficulties in using wafers with multilayer structures. This paper describes an improved method for end-point detection of wafers with multilayer structures which consist of 160 +/- 10 nm SiN/20 nm SiO2 /Si. The improvement has been done in the following way. (1) Variation of the film thickness in the multilayer structure was found to cause shifts of the end point and CD. (2) By choosing the optimum wavelength, the CD shift was minimized for the variation of the film thickness. (3) The optimum wavelength was determined from the reflectivity measurement on the multilayer structure. (4) Improved CD accuracy of +/- 0.02 micrometers was attained for a SiN film thickness varying from -10 nm to +10 nm relative to an average value by choosing the optimum wavelength.

This work considers the modeling and control of positive optical photoresist development. The control effort is applied to this final process step in order to compensate for deviations in previous irreversible steps. A new approach to line-width control is studied. State estimation, combining interferometric development measurements with a process model, is used to calculate the time at which developer breakthrough to the substrate occurs. Parameter identification is used to determine any sample-to-sample changes in the process model. The optimal control policy is shown to be bang-bang with switching at the final time. A model describing development after breakthrough is derived and used to calculate the final developer shutoff time, based upon an on-line identified development process parameter. The policy is tested experimentally on different batches of the same resist type. Results indicate that accurate control is achieved, despite changes in environmental conditions and unexpected transient disturbances.

Design and operation of an on-line Track Development Rate Monitor (TDRM) for track development systems are described. Dissolution data, measured using this equipment, for spray and puddle development processes is presented and compared to those derived from a conventional immersion DRM. Immersion data has traditionally been used to model all development. The validity of this is discussed. Also presented is an off-line technique for evaluating dissolution rates which utilize no specialized DRM equipment. The dissolution rates as measured by this technique are compared with those obtained from the TDRM/DRM methods. Simulations using all the calculated dissolution parameters are compared with SEM cross sections so that a practical evaluation of the various techniques can be made.

As the microelectronics industry strives to achieve smaller device design geometries, control of linewidth, or critical dimension (CD), becomes increasingly important. Currently, CD uniformity is controlled by exposing large numbers of samples for a fixed exposure time which is determined in advance by calibration techniques. This type of control does not accommodate variations in optical properties of the wafers that may occur during manufacturing. In this work, a relationship is demonstrated between the intensity of light diffracted from a latent image consisting of a periodic pattern in the undeveloped photoresist and the amount of energy absorbed by the resist material (the exposure dose). This relationship is used to simulate exposure dose control of photoresist on surfaces which have different optical properties chosen to represent surfaces typical of those found in operating process lines. Samples include a variety of photoresist materials and substrates with a wide variety of optical properties. The optical properties of the substrates were deliberately varied to determine the effect of these properties on CD (in the presence and absence of an exposure monitor) during lithography. It was observed that linewidth uniformity of the developed photoresist can be greatly improved when the intensity of diffracted light from the latent image is used to control the exposure dose. Diffraction from the latent image grating structures was modeled using rigorous coupled wave analysis. The modeling is used to predict the diffraction from a latent image as a function of the substrate optical properties and the parameters of the latent image (i.e., linewidth, sidewall angle). Good agreement is obtained between theoretical and experimental observations. Conversely, the inverse problem is solved in which the parameters of the diffracting structure (the latent image) are determined from a measurement of the diffracted power. Therefore, the diffracted power can be monitored for the purpose of determining when the latent image will produce the proper CD upon development.

Design and implementation of a multifunctional 'golden standard' wafer which facilitates easy and rapid measurement of the most important optical stepper parameters using commercially available metrology tools and software are described. In this design a stepped array of reference marks is permanently etched into a silicon oxide film on a silicon wafer. The most important feature of this method is that a permanently etched pattern is used as a reference standard, which may be used to compare a number of steppers over a period of time. When it is required to perform a distortion measurement the wafer is simply coated with photoresist and a series of full-field exposures are exposed on the stepper to be measured. Using this method, the errors due to individual stepper stages are removed since each lens distortion measurement is made with respect to a common grid pattern. Furthermore this method provides useful information regarding the accuracy of the stepper alignment system. In addition the permanent alignment marks on the wafer permit the measurement of individual stepper stage matching, as well as stage orthogonality errors.

Measurement precision, especially the measurement offset of an automatic overlay measurement technique, was studied for application to sub-half micron device manufacturing. Experimental data showed that the measurement offset depended on the cross-section structure rather than the reflectivity or the roughness of the overlay marks. Dependence of the measurement offset upon equipment factors such as the incident angle of illumination was also studied. This paper also shows measurement offsets on critical levels of the sub-half micron device manufacturing.

An experimental assessment of wafer flatness for 150 mm P/P+ epitaxial silicon wafers is presented to illustrate the use of graphical and analytical statistical techniques to both characterize and determine methods to reduce the variation in wafer flatness. This approach is applied to the chucked Local Total Indicator Reading (LTIR) site least squares front-reference plane for 354 P/P+ wafers and 144 P-type wafers from three suppliers measured over a range of site sizes (15 mm X 15 mm, 20 mm X 20 mm, 25 mm X 25 mm, and 20 mm X 30 mm). The objective is to ascertain the extent to which the lens total indicator range budget (taken as 1.1 micrometers as a figure of merit), is utilized by the silicon material. After taking into account estimates of the circuit topography and the lithographic machine detractors, an upper bound of 0.52 micrometers for the silicon LTIR was determined by an RMS analysis and a lower bound of 0.30 micrometers was taken as an approximation to a linear analysis. Histogram, cumulative frequency plots, and boxplot analysis for the sites on the wafers are presented. The importance of correcting for 'abnormal' site locations on the gauge chuck is noted. The percent variance components of flatness within a wafer, wafer to wafer in an epi run, epi run to epi run, polish lot to polish lot, and shipment to shipment is also presented. The source of the epi wafer percent variability appears to reside in the polished wafer substrate rather than be intrinsic to the epi process per se by comparing 'before' and 'after' epi wafer LTIR data.

The characterization of lithography tool performance during evaluation and in the production tool line is traditionally a time-consuming operation. The work presented here describes a technique which uses scattering of an optical probe from latent images in exposed, undeveloped resist to make rapid and reproducible measurements of a number of optical projection tool printing characteristics. Lens characteristics such as field curvature, astigmatism and coma as well as machine parameters such as column tip, focus and standard exposure dose may be measured. In addition, resolution and defocus sensitivity characteristics may be observed. Results have been obtained on exposure tools operating at various wavelengths and numerical apertures. The speed and accuracy of Latent Image Metrology (LIM) has enabled precise determination of exposure and barometric pressure induced focus variations. Changes in the field curvature, astigmatism and other imaging properties of lenses have been observed when they are subjected to high exposure doses.

Phase-shifted patterns (alternating, 90-degree, and chromeless) have been incorporated into a reticle layout, fabricated with a MEBESR III system, and evaluated experimentally at 365 nm using steppers with numerical aperture (NA) ranging from 0.4 to 0.48 and partial coherence ranging from 0.38 to 0.62. Test circuit layouts simulate actual circuit designs with critical dimensions ranging from 0.2 micrometers to 1.2 micrometers . These results, combined with experimental measurement of layer to layer registration and aerial image simulations, provide a first-order assessment of e-beam lithography requirements to support phase-shift mask technology.

One of the problems in applying the phase-shifting method in the positive resist process is the resist bridge generated at the phase-shifter edge. This problem has occurred in the past because the light intensity decreased to zero due to the interference at the phase-shifter edge. In order to solve this problem, we propose a new phase-shifting mask structure containing an intermediate phase-shifter. This intermediate phase-shifter will change the phase of the light by 90 degrees and will be placed at a peripheral edge of the conventional phase-shifter on the transparent substrate. The effect of this mask structure is demonstrated. A 0.3 micrometers lines and spaces pattern is successfully resolved without resist bridge, and the DOF at a 0.35 micrometers lines and spaces pattern is 1.2 micrometers wide. It is also demonstrated that this mask structure is effective on patterns such as LOCOS.

Rigorous simulation of electromagnetic diffraction with TEMPEST is used to explore the impact of edge profiles, phase-shifting materials, protective coatings and reflective masks on projection printed image quality. The TEMPEST massively-parallel finite-difference time-domain scattering analysis program has been extended to generate diffraction efficiencies for transmitted as well as reflected fields. The mask materials and geometries are input through specifying turning points along the polygonal boundaries of the chrome, phase- shifting materials, overcoating, etc. Plane waves in the TE orientation are then used to illuminate the mask, and a postprocessor is used synthesize the image from diffracted fields. To identify problematic situations and survey interesting technology approaches, a variety of proto-typical mask geometries are considered. The fundamental problems of the effects of rounding of chrome edges are investigated using optical parameters at 248 nm. Phenomena which might contribute to improved image quality with overcoating are studied using planar, conformal and inhomogeneous coating models at i-line. Effects from large (0.365 nm) vertical and overcut edges in phase-shifting layers are also explored. Finally, for the reflective mask technology, the effects of mask edge angles and the use of built-in materials-based phase-shifting are explored.

Computer simulations and i-line phase shift lithography experiments with programmed 5X phase shift reticle defects were used to investigate the effect of opaque and phase-shift layer defects on sub-half-micron lines. Both the simulations and the experiments show that defects in the phase shift layer print larger than corresponding opaque defects, with 0.3-0.4 micrometers defects affecting sub-half-micron critical dimensions by more than the allowable 10%. Inspection of programmed phase shift defects with a prototype mask inspection system confirmed that the system finds the 0.3-0.4 micrometers phase shift defects critical to sub-half-micron lithography.

Decreasing dimensions in the processing and manufacture of integrated circuits (ICs) has stimulated interest in ultra-high resolution measuring technologies. The Atomic Force Microscope (AFM), which is now available commercially, offers three-dimensional surface measurement capability from angstroms to over 100 microns, the ability to image insulators directly without coating, and minimal sample preparation. These features indicate strong potential for applications in IC related inspection, process engineering, failure analysis and reliability, particularly as ICs move toward submicron geometries.

Identifying particulate sources at an intermediate stage in the VLSI manufacturing process is frequently indispensable for taking the appropriate corrective action, and thereby enhancing device yield. To rapidly locate and analyze both organic and inorganic particles on a large silicon wafer, the authors have developed an automated wafer-surface particle detection and identification system that consists of a laser-scanning wafer inspection unit, an SEM/EDX unit, a microfluorescence spectroscopy unit, and a computing unit for file conversion and XY-coordinate transformation. Examples of successful applications of a particle inspection and analysis program using this new system in VLSI wafer processing are presented. With emphasis on photolithography and the subsequent processes such as ion implantation and dry etching.

Techniques commonly used to determine silicon penetration depth during deep UV surface imaging lithography are compared to a method referred to as plasma etch 'staining.' This methodology is described in detail and the results compared and correlated to Rutherford Backscattering Spectroscopy (RBS) and ellipsometric (film swelling) measurements. Effects of the staining parameters on the resulting silicon depth are also discussed.

There are several methods for correlating thin film thickness measurement equipment. The most commonly used method is to program in an offset that gets systematically added to or subtracted from the thickness calculation. Measurement errors can occur using this 'fudge factor' method because the relationship between the reference thicknesses and the measured thicknesses is not, in general, a fixed constant or simple function. A better method for correlating thin film thickness measurement equipment is to tune the instruments to the process by refining the internal optical constants.

The authors investigate the possibility of using monochromatic ellipsometry to measure the thickness of amorphous silicon (a-Si), titanium (Ti) and a-Si/Ti overlayers on crystalline silicon (c-Si) substrates. Accurate thickness control of these layer structures are very important in titanium silicide straps formation which are often used as interconnects in integrated circuits. Films with different layer structures, a-Si/c-Si, Ti/c-Si and a-Si/Ti/c-Si are analyzed using a monochromatic ellipsometer operated at 6328 angstroms. The thickness of the desired layer is derived from the theory of ellipsometry implemented on a home-made computer program by knowing the optical constants of the layers from literature. The results illustrate that it is possible to measure the aforementioned layer thicknesses using a commercially available monochromatic ellipsometer. However, the results are sensitive to such factors as native oxide thickness, inaccuracy of the angle of incidence during the experiment, and uncertainty in the layer optical constants. We illustrate the aforementioned sensitivities. Simulation results and a comparison of measurements with theory are discussed.

The continuous reduction in the dimensions of IC features has resulted in the need for higher resolution metrology tools. Low voltage scanning electron microscopy has the advantages of offering high resolution and a large depth of field. In addition, its non-destructive nature yields itself to in-process metrology without sacrifice of product. In this investigation, an advanced field emission scanning electron microscope is employed in the determination of optimal conditions for the inspection and measurement of common device- grade thin films. Polysilicon and silicon nitride films are characterized using several statistical tools. A factorial design experiment is used to narrow the range of effective SEM conditions. Subsequent measurement batteries will address the stability of the film-SEM interaction over a series of measurements, and resolve the fine degrees in condition performance. A treatment of the long-term charging and dimensional growth effects is also included.

As submicron VLSI production becomes standard, metrology tools tend to rely on the intrinsic high resolution of the e-beam. Interactions of electrons with the insulating materials typical to the IC industry generate a charging phenomena. In this paper charging types are classified and modeled. Since charging can have severe effects on the accuracy, precision and stability of metrology tools, proper countermeasures must be taken. This is especially true as these systems become automated, in order to obtain correct measurements without manual intervention. The authors discuss diagnostic techniques which indicate the type of charging and the appropriate countermeasure. The paper is organized with, first, a discussion of the measurement process: relating the waveform to the real feature. Then the basic physics of charging is discussed with order of magnitude calculation. The influence of charging on measurement is then detailed relative to the three basic metrology qualifiers: accuracy, precision and stability. Practical methods for avoiding the effects of charging are included.

The three conventional techniques--optical, low voltage scanning electron microscopy (LVSEM), and electrical linewidth measurement--continue to be employed, but each technique has unique applications, problems, and limitations. In this paper these techniques are investigated for submicron linewidth metrology. A great deal of emphasis is placed on the calibration of these tools and the potential for problems associated with the tools.

A new CD measuring tool based on confocal scanning technology, Carl Zeiss Axioscan, is described. The advantages of having a great variability in choosing the illumination is demonstrated with several examples. Selection of deep UV illumination increases the lateral and height resolution of the system. Matching the illumination to the optical constants of the materials to be measured enhances the reproducibility and the accuracy of the measurement. In the description of the system, it is shown how the confocal scanning technique can be implemented with conventional light sources, and how this increases the flexibility in illumination conditions.

A new measuring method was developed to optimize the submicron CD measurements with a conventional optical microscope system. An optimum combination of the inspection system optical parameters is used to accurately and precisely measure each feature of interest. Such combinations are considered to define the 'operating points' for the new measuring method. Also, the slope of the logarithm of the image intensity profile was determined to be an appropriate metric of aerial image quality in order to predict the operating points number and their placement. This paper discusses the experimental results obtained in measuring 0.75 micrometers isolated spaces with the proposed method and the construction of the operating points for this feature, with the measured linewidth data.

STM (Scanning Tunneling Microscopy) and its variants, collectively called SXM (not including electron microscopy), are being increasingly used in electronics research and industry to study and inspect semiconductor surfaces. While it is possible to obtain quite high depth resolutions, lateral spatial resolution is chiefly determined by probe size and shape which is typically much larger than the depth resolution. Thus, an SXM image is not a reflection of the true surface shape but rather a 'convolution' of the surface and probe shapes. This paper reviews the theoretical and experimental work done in reconstructing surface shape from SXM images. The authors present a computational model of the SXM imaging process that encompasses previous models and show that the imaging process (convolution) is essentially a nonlinear operation and can be approximated mathematically by a morphological dilation between the surface and probe shape. The authors address the problem of inverting this process to estimate the true surface shape. A general method is developed by which a surface can be reconstructed from a composition of SXM images produced by different scanning probes. A multi-resolution version of this composite method is then described using a set of multiscaled probes that can recursively and efficiently reconstruct the entire surface as though it had been scanned entirely by the smallest probe. Some other useful and interesting results of our SXM imaging model are presented. The authors conclude by discussing the theoretical and practical importance of their computational model of SXM imaging process and directions for future work.

Optical instruments for submicron linewidth measurement are limited by the resolution associated with the wavelength of light. By measuring phase, rather than intensity, the lateral resolution of the system can be improved. This paper describes a system based upon an interferometric optical profiler using a Linnik interferometer and narrow-band illumination. The system measures surface height directly using phase-measurement interferometry techniques. Three-dimensional maps of surface structure over 1024 X 1024 pixels are produced. The system includes an 80486-based AT-compatible computer and a Videk Megaplus camera. The Videk Megaplus camera has 1320 X 1030 active square pixels spaced at 6.8 micrometers intervals. It outputs an 8-bit digital signal which is interfaced to a framegrabber, and intensity is displayed live on a high resolution monitor. The incoherent optical resolution of the system is 0.34 micrometers , and the detector samples the surface every 0.034 micrometers . It has been found that features as small as 0.5 micrometers can easily be measured. Two calibration standards traceable to NIST with feature sizes on the order of 1 micrometers have been used to calibrate the lateral dimensions of the system. Results of measurements of photoresist lines on silicon with 0.5 micrometers lines are presented.

As the demand increases for smaller, more powerful new products, design engineers are pressed to increase component densities while simultaneously reducing the size of interconnects. Almost every new product contains more solder joints per square inch than the previous one. New quad flatpack and TAB designs as small as 25 micron leads on 50 micron centers are in the prototype stage. Direct 'chip-on-board' (COB) components placed on a grid of solder bumps with a diameter of 75 microns and a grid of 200 microns are routinely being produced. Future plans include designs with a 25 micron diameter on 50 micron centers which are currently in development. Devices consisting of chips stacked upon chips and interconnected with solder or tungsten wires are increasingly included in new designs. Manufacturers have also begun to produce assemblies on very thin circuit boards with components on both sides. Several technologies have been applied in an effort to provide solder paste and post- reflow inspection. X-ray inspection has proven most effective at determining component placement and solder joint integrity. With its ability to pass freely through circuit board materials and extract detailed structural information from hidden and visible solder joints, the x-ray has proven more adept at assembled board inspection than other automated methods such as laser, ultrasonic, thermal and camera-based systems. This paper addresses the inspection and process control of ultra-thin boards with ultra fine pitch interconnects using x-ray laminography. In addition, the advantages and disadvantages of integrating various fine pitch technologies into the circuit board assembly process are reviewed.

An expert system called SEIB (sustaining-engineer-in-a-box) has been developed to allow routine lithography troubleshooting to be performed by manufacturing personnel, reducing their dependence on sustaining engineers. SEIB is a multi-media, multi-interface expert system, capable of utilizing interactive video images, PC-based information, and database information contained within a mainframe computer-aided manufacturing (CAM) system. SEIB utilizes a custom expert system contained within a DVIR interactive video platform. This allows for quick resolution of problems without engineering intervention. Because of its design, the system can be easily adapted to any functional area in the fabrication facility.

Adaptive control techniques, with their capability for providing satisfactory control even when the process changes with time, are promising candidates for dealing with common problems encountered in photolithography processing such as batch-to-batch variations in resist properties, inconsistencies in resist curing, etc. In this paper an adaptive control strategy for the photolithography process is proposed and evaluated. The design utilizes a reduced-order lithography model, an on-line parameter estimator, and a nonlinear model-inversion controller (NMIC). The width of the printed resist lines--a crucial output of photolithography--is controlled by automatically adjusting the exposure energy. In the calculation of the appropriate exposure adjustment, the controller uses both measured critical dimensions as well as estimated values produced by the process model. The control system is capable of tracking changes in the photolithography process by automatic updating of key model parameters as the process evolves in time. Simulation studies of the closed-loop adaptive control strategy using the PROLITH simulation package to represent the lithography process demonstrate the feasibility of this approach.

VLSI design rules require existing LSI design rules to extend to sub-micron and half-micron geometries. Using high resolution resist and 5X stepper (G- line) technology along with a Post-Exposure Bake (PEB) is a common method to improve the resolution. The PEB drives out residual photoresist solvents which can interfere with the develop process, resulting in CD variations. PEB strongly influences CD variations. The authors consider the following PEB parameters in this CD improvement study: (1) altering the PEB temperature, (2) altering the PEB time, and (3) altering the queuing time between PEB and cool prior to develop. The process characterization data includes critical dimension data for 0.8 micrometers lines, including proximity effects data on four high-resolution photoresists.

Conventionally, highly structured patterns or targets are placed on wafers to facilitate alignment and highly repeatable positioning of the wafer. The authors describe a micro-positioning scheme in which such patterns are replaced either by the device pattern on the front of the wafer, or the ground surface on the rear of the wafer. When illuminated by a visible or near-IR beam of light, both the patterned and the unfinished (i.e., diffusely ground silicon) sides of a wafer scatter prodigious amounts of light. By collecting portions of this light through two apertures and then measuring the phase of their mutual interference, the position of the surface wafer can be repeatedly and unambiguously established with a precision of 1 micro-inch (25 nm) over a range on the order of d equals (lambda) /A, where (lambda) is the wavelength of the light and A is the angle subtended by the apertures at the wafer. Typical values for (lambda) and A are 0.8 micron and 0.1 radian, respectively, in which case d approximately equals 8 microns. This means that if a standard stage can be used to position the wafer to within +/- 4 microns, then the interferometric sensor described here can be used to refine its position to within 25 nm. The ultimate resolution, (Delta) xmin, with which this position can be reestablished is equal to (lambda) /(A'SNR) where SNR is the signal-to- noise ratio of the interference signal. For example, with d equals 8 micrometers and a signal-to-noise ratio of 400, (Delta) xmin approximately equals 25 nm, or just under a micro-inch. The direction of the sensitivity vector direction is defined by the relative orientation of the two apertures, and by the associated optics. By using two measurement locations, angular orientation of the wafer about an axis normal to its surface can also be monitored. Because the measurement is inherently based on interferometric phase, rather than the amplitude, it is highly tolerant to variations in surface reflectivity and/or illumination level. The authors describe the theoretical basis for this measurement technique and present results demonstrating repositioning to better than 50 nm using both the patterned face of a wafer and its diffusely reflecting back side.

A combination of metrology techniques was employed to fine tune the wavelength setting of a 248 nm excimer laser stepper to optimize performance. Scanning electron microscopy was used to document local resolution, proximity effects, and astigmatism, while GCA SMARTSETR and electrical resistance techniques were used to examine full field effects. Using the combined metrology methodologies, the authors documented the decrease in proximity effect, improvement in resolution, and increase in absolute lens distortion with negative shifts in laser wavelength setting, with a slight differential in the setting required to minimize horizontal versus vertical proximity effect and astigmatism. A wavelength offset of -2.3 angstroms from the nominal stepper setup wavelength was determined to be the best operating wavelength for these applications.

Polysilicon over gate oxide thickness measurements are one of the most important thickness measurements in today's semiconductor manufacturing industry. Polysilicon thickness variations affect the film's ability to perform efficiently in controlling implant distribution. Further, polysilicon thickness variations can cause variations in electrical characteristics like IDSAT. Theory predicts that the inherent physical properties of the polysilicon will limit the effectiveness of optical thin film thickness measurement instruments, especially when measuring over very thin oxides. An experiment was performed to determine the effect of underlying oxide thickness on optical polysilicon over oxide thickness measurements.

Visible and ultraviolet light reflectometry provides a fast, convenient, and nondestructive method of characterizing multilayer film structures that include polycrystalline silicon. Reflectance measurements of silicon wafers containing such films have provided information as to the roughness of the poly surface, the thickness of the films, and the optical properties of the poly.

The authors report on the angle resolved light scattering characteristics of individual polystyrene spheres on three silicon surfaces. A He-Ne laser (632.8 nm) focused to a 15 micrometers 1/e2 diameter was employed to illuminate 0.804 micrometers diameter spheres on optically smooth ((sigma) >= (lambda) ) silicon surfaces: monocrystalline silicon (bare silicon), polycrystalline silicon (polysilicon), and roughened silicon ('black silicon'). These surfaces provided a roughness spectrum ranging between the smooth, virtually featureless surface of the bare silicon to one of dense, very coarse needle- like features on the black silicon. Scattering was measured as a function of incident beam polarization for incident angles of 30, 45, and 75.3 degrees (Brewster's angle). Experimental measurements show that the beam incident angle and polarization are important factors controlling substrate background scatter. The substrate influences sphere detectability in two ways: first, by directly scattering incident radiation into the detector and second, by reflecting a portion of the forward scattered light originating from the sphere. The results obtained are qualitatively explained with reference to the silicon surface reflectance which varies significantly as beam incident angle and polarization are changed. As surface roughness approached a value comparable to the sphere size, detectability diminished; that is, the measured cross sections were lower under these conditions. Surface roughness not only added to the background signal, but also reduced the amount of energy scattered by the sphere reaching the detector. A modification to Lorenz-Mie theory is introduced to explain the experimental findings. The approach taken is to calculate the scattering components for a sphere in free space, then attenuate those components which are reflected from the substrate into the detector. The computation includes the variation in substrate reflectivity as the ray incident angle and polarization change. The assumptions and validity of this approach are discussed, as well as future possible improvements to the model.

Reducing defect density to an acceptable level is one of the most challenging problems in manufacturing 16 Mbit DRAMs. To implement an effective defect elimination strategy, it is necessary to have both a defect inspection system that can find all critical defects on the wafer and a strong defect reduction methodology. The optimum system would provide the highest sensitivity along with sufficient inspection speed to fit a majority of inspection applications. This paper describes a new system, the KLA 2110, designed to meet production requirements for speed, sensitivity, and ease of use. The system provides 0.25 micrometers sensitivity with image processing rates 100 times faster than previous generation digital image processing technology.

Patterned wafer defect inspection has evolved from manual to automated wafer inspection equipment. The success of certain applications (and resulting yield improvement) using automated wafer inspection equipment is well documented. Automated wafer inspection equipment currently in use in the semiconductor industry has primarily been utilized in engineering analysis applications. Advances in inspection technology now allow for automated wafer inspection to become a more integral part of the manufacturing strategy. As automated wafer inspection becomes more widespread in the manufacturing line, data analysis issues specific to automated wafer inspection arise. This paper describes general patterned wafer inspection applications and associated requirements in terms of ensuring the quality of data generated. Techniques to help the user more efficiently gather and use defect data are presented. Key equipment characteristics needed for each application are then reviewed.

Automated defect detection using various techniques (including holography, digital image processing and particle detection) has provided process engineers with defect distributions and densities for all process levels. However, defect detection systems alone cannot differentiate between killer and nuisance defects, and can only give an indication of the potential yield reducing problems. The location of the defect with respect to the process level is important in determining the impact on the final device. Recent techniques have been developed to automatically separate killer defects from nuisance defects. This paper show how this technique is used to isolate the yield reducing defects. The method used shows how results from electrical test can be correlated with defect inspection results.

A wafer metrology 'round robin' has been completed comparing linewidth measurements from several different companies and different measurement tools and technologies. The project has been conducted under the auspices of SEMI by members of SEMI's Metrology Committee. The goals of the program were: (1) determine the range of critical dimension values measured across the United States, (2) test the newly formulated SEMI linewidth patterns, (3) assess the effect of calibration differences of the measured values.

The use of statistically designed experiments for optimizing photoresist processes in manufacturing environments provides a powerful method of establishing a robust process. Unfortunately, there is an abundance of experimental designs, approaches, and analysis techniques available to the experimenter. A sequential approach of experimental design will be outlined which can logically and efficiently take the lithography process engineer from the initial problem objective(s) to the final optimized process. The test vehicle used to demonstrate this approach is a novel negative tone photoresist which is based on a positive type chemical formulation with a novolac-bound isourea. Transformation of a multivariable response to simultaneously locate optima from seven original design variables is employed. Minimization of film loss, maximization of process latitude, and optimization of resist resolution to a fixed target are concurrently achieved through manipulation of variables including bake temperatures, exposure energies, and developer normality. With the variables and responses specified, the first stage of the sequential approach is an initial screening experiment based on an extended Morris Mitchell design with center point replicates. This efficient design minimizes experimental trials while limiting confounding of interactions between 1st and 2nd order coefficients. Using the technique of steepest ascent to select a region of optimal response values, a second experiment is performed using a factorial design. Only those variables showing significance levels greater than 95% are carried forward from the screening analysis. Finally, a response surface based on a full second order quadratic response surface model (RSM) is demonstrated. Model validity, confidence intervals and significance levels are investigated. Three different process options are derived from the RSM process model and are experimentally verified.