A rigorous numerical technique to calculate images of optically thick one-dimensional objects as formed by an optical scanning microscope is presented. The geometry of the selected system allows us to simulate microscopes in various modes of image formation (coherent, partially coherent, bright-field, dark-field, and confocal). We consider the cases of s- and p- polarized illumination. The samples consist of dielectric film structures deposited on perfectly conducting flat substrates, and are characterized by the surface profile, the film thickness, and the complex refractive index of the dielectric. Examples of bright-field coherent images calculated with this technique are shown.

Real-time process monitoring has been a primary concern of the semiconductor industry for a number of years. In an attempt to provide higher yield and performance it has become accepted that the monitoring of the etch step is critical. This is the primary motivation for the development of a real time process monitor with particular attention paid to wafer monitoring at the surface. To this end, the use of the results of diffraction, in particular, diffraction from 3-dimensional gratings is proving to be a viable technique for real time, process monitoring. Thus, the primary focus of this paper is to present the progress made toward the development of a non destructive, real time, optical metrology based system for direct wafer monitoring, utilizing the results of diffraction from a grating structure. The results of experiment and simulation will also be discussed.

An investigation is being carried out to determine the ability of three methods of linewidth metrology to measure the dimensions of features of less than 0.5 micrometers . The three methods are transmitted-light optical microscopy, electrical test structure, and scanning electron microscopy (SEM). To permit the inclusion of transmitted-light optical microscopy in this investigation, 100-nm thick Ti films were patterned using normal VLSI processing techniques on a 150-mm diameter quartz wafer. The cross-bridge resistor test structure was used since this structure has been widely used in industry and it allows the results from all three metrological techniques to be compared. The design bridge widths of the test structures range from 0.4 micrometers to 1.0 micrometers . The results of these measurements show systematic and uniform offsets between the different techniques. In this paper we discuss the different techniques and describe the observed results.

Using etched quartz as a phase shifter for i-line phase shift masks requires an etched depth of 385 nm referenced from the quartz surface. Recent work shows a direct relationship of focus offset as a function of the phase angle as it deviates from 180 degree(s). Knowing the etched depth and phase in transmission becomes critical to the production and verification of these masks. A method of interferometrically evaluating a phase shift mask is proposed. The method calculates the phase shift from the surface profile of the etched shifter assuming that a good optical surface is maintained on the unetched regions. This surface measurement method possesses high spatial resolution at the expense of only knowing the amount of phase shift from the profile of the etched quartz shifter. Correlations between this method and mechanical stylus measurements establish the validity and advantages of this measurement technique.

Using etched quartz as a phase shifter for i-line phase shift masks requires an etched depth of 385 nm referenced from the quartz surface. Recent work shows a direct relationship of focus offset as a function of the phase angle as it deviates from 180 degree(s). Knowing the etched depth and phase in transmission becomes critical to the production and verification of these masks. A method of interferometrically evaluating a phase shift mask is proposed. The method calculates the phase shift from the surface profile imaging system. Results show that errors of less than 5% are obtained for features as small as 1/6th of the optical resolution of the system and less than 2% for features as small as 1/2 of the resolution. When noise is included in the modeling, errors of 5% are obtained for S/N of 10 and less than 2% for S/N of 100.

Quantitative methods are developed to use optical scatter to measure the critical dimensions of gratings etched into bulk Si and developed photoresist patterns on silicon substrates. Previous work either classified microstructures qualitatively or employed a 'chi-by-eye' method to find that structures were similar or dissimilar. A single detector scanning scatterometer is used to measure large 32 micrometers pitch structures while another instrument that varies the angle of incidence and tracks diffracted orders via the grating equation is used to measure 2 micrometers pitch structures. A rigorous coupled wave light scatter model is used to simulate diffraction from a set of test wafers. Partial least squares and neural network analysis techniques are then employed to use correlations between the simulated diffraction and the critical dimensions of the modeled structures to produce a capability to measure the critical dimensions from scattered light measurements. The marriage of rigorous coupled wave diffraction modeling and optical scatterometry directly addresses the needs of the industry for a rapid and nondestructive metrology tool.

In this paper we discuss an optical metrology technique for the determination of optimum lithography parameters through an interrogation of the latent image. This technique, called the Lithography Process Monitor (LPM), involves illuminating a latent image grating with a laser beam. The intensity of the orders diffracted from the grating has been shown to be directly related to the photoactive compound (PAC) concentration profile, and consequently, to the profile of the developed resist. We have developed a method of modeling the intensity in the diffracted orders by using lithography simulation software in conjunction with rigorous coupled wave diffraction analysis. Experiments have been conducted with both positive and negative resist. In addition, we have been able to determine the 'absolute' location of the top of the photoresist with respect to the stepper focal reference and determine film thickness variations on the wafer.

Cost of Ownership (CoO) models are increasingly a part of the semiconductor equipment evaluation and selection process. These models enable semiconductor manufacturers and equipment suppliers to quantify a system in terms of dollars per wafer. Because of the complex nature of the semiconductor manufacturing process, there are several key attributes that must be considered in order to accurately reflect the true 'cost of ownership'. While most CoO work to date has been applied to production equipment, the need to understand cost of ownership for inspection and metrology equipment presents unique challenges. Critical parameters such as detection sensitivity as a function of size and type of defect are not included in current CoO models yet are, without question, major factors in the technical evaluation process and life-cycle cost. This paper illustrates the relationship between these parameters, as components of the alpha and beta risk, and cost of ownership.

As photolithography technology advances into submicron regime, the requirement for alignment accuracy also becomes much tighter. The alignment accuracy is a function of the strength of the alignment signal. Therefore, a detailed alignment signal intensity simulation for 0.8 micrometers EPROM poly-1 layer on ASM stepper was done based on the process of record in the fab to reduce misalignment and improve die yield. Oxide thickness variation did not have significant impact on the alignment signal intensity. However, poly-1 thickness was the most important parameter to affect optical alignments. The real alignment intensity data versus resist thickness on production wafers was collected and it showed good agreement with the simulated results. Similar results were obtained for ONO dielectric layer at a different fab.

Current generation of advanced IC's require sub-half-micron-resolution photolithography over large exposure fields, with device overlay tolerances of less than 100 nanometers. Achieving this performance in high-volume manufacturing will challenge the focus and overlay control capabilities of optical reduction steppers. This paper presents new metrological approaches to achieving the required focus and overlay control performance. A latent image focus measurement technique is described, which has been used extensively to characterize die levelling performance. To improve overlay capability on back end levels (especially metal), a combined bright field/dark field alignment system has been developed. Data on alignment performance, and optimization of alignment mark design for bright field, will be presented.

The work described in this publication details a procedure to determine the capability of the photochemical system utilizing a combination sub-develop and sub-exposure technique. The development of the test procedure concentrated on the ability to minimize the degree of random noise associated with the given response function. A reduction in the noise level allowed for a maximum signal and thus, the ability to detect slight variations within the photochemical constituent. To accomplish this task with application into high volume manufacturing, a two-fold approach was initiated. The first approach utilized a multilevel statistical experimental design. Through the use of RSM, a model was developed to statistically interpret the response as a function of the input variables. The second approach utilized a series of 'single cell' experiments.

Excimer laser lithography combined with chemically amplified resists offers a viable approach to lithography at 0.5 micrometers and below. APEX-E, a positive tone deep-uv resist used in conjunction with a 0.44 NA excimer laser stepper is capable of 0.35 micrometers resolution. To improve the process window for this resist while reducing the performance variability, the Taguchi method of quality control was employed. The baseline process for APEX-E was characterized, then subsequently used as a comparison to the optimized process as suggested by the Taguchi experiments.

Lithographic equipment for processes with 0.35 micrometers design rules is becoming available. One of the critical questions is which imaging technology will be used to define the required small feature size. Another question, frequently underestimated, is how to achieve the required overlay requirement of 120 nm between randomly chosen machines. A general methodology to maintain the matched machine overlay has been reported earlier. However, since it is believed that a mix of lithographic tools will be used at 0.35 micrometers , namely DUV and i-line, the task of multiple machine matching is even more challenging: the overlay has to be maintained between two different types of steppers. The lens matching required--also between DUV and i- line lenses--is found to be 70 nm. In this paper we will report on the matched overlay performance of 14 machines, resulting in data from 52 matched random pairs of both DUV and i-line ASML steppers. The matching methodology is optimized to maintain optimum performance between random pairs of machines. This includes the use of reference wafers, matching simulation and stepper selection. Furthermore, we investigated different matching mathematics and their influence on matching performance. We show that the 120 nm overlay requirement can be achieved using multiple machines in a production environment. The paper concludes with a device overlay budget analysis.

The imaging performance of IBM's positive DUV resist is evaluated on SVGL Micrascan. Preliminary results on GCA excimer laser are included. Contamination effects are studied using 0.5 micron lines and spaces on Micrascan I using cross-sections at various delay times. The effects of delay times between different processing steps such as coat to expose and expose to PEB are studied. The performance is further evaluated by processing wafer lots on line for Micrascan I and II. Data on resolution, depth of focus and process latitude are presented for 0.35 and 0.5 micron geometries. The effects of PEB temperature variation are studied. Performance on different substrates (poly, metals and contacts) is evaluated.

We have examined several metal films currently used in IC manufacturing and reported the results of reflectivity as a function of different processing parameters. The objective when characterizing a film to be used as an anti-reflective coating (ARC) is to locate and optimize the process window to achieve the minimum reflectance at the operating wavelength of a stepper. Reflectivity measurements as a function of thickness and process conditions, across broadband wavelengths, are presented. These results show the variation in minimum reflectance as a function of these variables. Various thicknesses of coherent titanium nitride (TiN) and titanium tungsten (TiW) films have been studied to understand the application of these films to interconnect metallization schemes as anti-reflective coatings. Additionally, a series of aluminum (Al) wafers are created to show the variation in measured absolute reflectivity as a function of process parameters. The aluminum deposition temperature was varied from 100 degree(s)C to 500 degree(s)C.

A new metrology method is described which is capable of extracting information about the flatness and performance of the vacuum chuck which holds the wafer in a photolithographic stepper. The method is based on a comparison of the wafer flatness as measured by the stepper and the wafer flatness as measured by a thickness-based flatness metrology system. The flatness datamaps for one wafer are subtracted using ensemble mathematics to yield a stepper chuck signature datamap. Signatures for several wafers are ensemble averaged to test for statistical validity. Stepper and flatness metrology data as well as statistical signatures are presented. Wafer and operational dependencies are discussed. The signature has potential use in the tracking of chuck wear and the feeding forward of flatness information to stepper control systems.

NIST is in the process of developing a new scanning electron microscope (SEM) magnification calibration reference standard useful at both high and low accelerating voltages. This standard will be useful for all applications to which the SEM is currently being used, but it has been specifically tailored to meet many of the particular needs of the semiconductor industry. A small number of test samples with the pattern were prepared on silicon substrates using electron beam lithography at the National Nanofabrication Facility at Cornell University. The structures were patterned in titanium/palladium with maximum nominal pitch structures of approximately 3000 micrometers scaling down to structures with minimum nominal pitch of 0.4 micrometer. Eighteen of these samples were sent out to a total of 35 university, research, semiconductor and other industrial laboratories in an interlaboratory study. The purpose of the study was to test the SEM instrumentation and to review the suitability of the sample design. The results of the analysis of the data obtained in this study are presented in this paper.

This paper discusses nonlinear behavior in SEM CD measurements stemming from the interaction of the edge detection algorithm and systematic changes in the appearance of the secondary electron signal. A first-order theory is developed which describes the effect of changes in apparent sidewall slope and baseline level on common edge detection algorithms. Specifically the theory predicts nonlinear behavior for linear approximation and threshold edge detection algorithms. Experimental verification of the effect is presented. Systematic increases in linewidth of 10% are frequently encountered in practice when making measurements in the sub 0.75 micrometers regime.

A novel atomic force microscope (AFM) is used to image a microlithographic sample. The AFM operates in the non destructive non-contact mode, uses glass tips as opposed to tungsten or silicon, and has an optical interferometric detection system. Its estimated lateral resolution is under 10 nanometers and much better in the z direction. A sample consisting of chrome features on quartz was produced for measurements using AFM and electric probe techniques. The features are single and grouped lines on the order of 1 micrometers incorporated into an electric probe pad layout. Dimensions of these features are determined from the AFM images by relating their sizes in pixels to the excursions of the scanners during the formation of the images. These results are compared with measurements obtained through electric probing techniques.

Several SEM inspection techniques for the process development of a 0.25 micrometers T-gate MESFET process have been successfully utilized. The transition of these techniques to a manufacturing process with SPC monitoring will also be discussed. The unique challenges of imaging several different Ebeam resists, and their interaction with the scanning electron microscope will also be discussed.

In a previous paper, a 2D atomic force probe metrology system was described. The system consists of a 2D atomic force probe used in attractive mode (non-contact) with a laser inter- ferometer system for measuring sample displacement with nanometer sensitivity. The force probe consists of a square cantilever and specially designed 2D tip. The tip and cantilever are vibrated in 2D and the shift in resonance as the surface approaches is used for surface sensing. The system is able to measure submicrometer lines and trenches with accuracy and precision at the nanometer level. This paper will discuss operational characteristics of the system and the application of 2D AFM metrology to calibration of optical and scanning electron microscope system.

In this work we present the results of a radically different approach to imaging of high-aspect ratio structures such as contact holes. Our approach utilizes two backscattered electron detection subsystems, one optimized for imaging at the top, like most SEM detectors, and another optimized for imaging at the base of submicrometer structures. These detection systems produce signals that can be combined in real-time to produce an image which resembles the 'extended focus' images obtained with confocal optical microscopes. Unlike confocal images, however, backscattered electron images have the inherent linearity and resolution characteristic of electron-beam technology. Backscattered electron imaging has been used to solve a number of vexing problems in monitoring semiconductor process. For example, contact hole measurement with secondary electrons has typically been done with a minimum or zero signal at the base of the structure, so that the measurement value obtained either has poor precision or is the result of extrapolation. In the case of backscattered electrons, the signal can have its maximum at the base of the structure, allowing high- precision measurement with no need for extrapolation. These results are supported by extensive Monte Carlo simulations.

We discovered the phenomena in scanning electron microscopy that the bottom of deep holes can be observed clearly when a high energy electron beam is irradiated. In order to utilize this phenomena, we developed a novel instrument that generates a scanning electron probe of 50 to 200 keV with TV-scan rate that enables us to inspect maximum 8 inches diameter wafers. In the past, we could inspect deep holes by observing cross sectional view of a hole milled with a focused ion beam (FIB). Using the novel instrument, we can easily inspect such holes without destroying samples. This observation does not depend on the specimen conditions, such as aspect ratio of holes and specimen charging that often occurs in the case of observing insulators. This instrument enables us to observe inside structures of semiconductor devices. Irradiation damage on semiconductor devices, such as threshold voltage and sub-threshold coefficient deviation of MOS transistors can be recovered, except for the irradiated area by means of hydrogen annealing of 30 minutes with temperature of 450 degree(s)C.

We report here survey results on pattern dimension measurement of masks and wafers. The SEMI Standard Metrology Committee in Japan have carried out a survey from June to September in 1992. The target of this survey consists of semiconductor device makers, inspection equipment makers, material makers, and academic institutes. We have got replies for ten questions from 44 respondents, and found some interesting results. E(Electron)-beam and optical equipment are mainly used in the R&D and the mass production lines respectively. In appears that the e-beam equipment is gradually replacing the optical equipment as a tool to measure finer pattern dimensions, however the e-beam equipment still has problems, like the measurement of contact-holes, or the assurance of the measurement accuracy which corresponds to the expected patterns' dimension. On the other hand, more than 2/3 of respondents have replied that they would like to standardize the measurement method itself in order to transfer process technologies from R&D to the mass production lines, or to make more flexible production lines, and about 85% of them have replied concerning the necessity of measuring finer pattern dimension absolutely to keep a common scale.

As the depth of focus of optical steppers grows smaller, it becomes more important to determine the position of best focus accurately and quickly. This paper describes the use of phase-shifted mask technology to form a focus vernier: a phase pattern on the stepper reticle which, when imaged in resist, can give both the magnitude and the direction of the focus error. In this, the focus vernier structure is analogous to 3overlay verniers. Thus the determination of focus error can be treated as an alignment problem in the z-axis. This technique is an improvement on previous schemes for the determination of best focus from resist images as it can indicate both the magnitude of the error and its direction in a single exposure.

As device densities continue to increase, lithographic requirements will continue to expand through the 90's. Although progress is being made in the resolution requirements in the 0.5 micron regime, resolution is not the only critical issue in submicron lithography. Interlevel alignment, or overlay registration, is also an area of vital concern. Various optical instruments are currently available to automatically determine interlevel alignment of photomask layers in integrated circuit manufacturing. It will be shown that the accuracy of those alignment instruments that use an optical column can be affected by mark design and the magnification in the optical column. The mathematical relationship between the alignment structure layout and the magnification of the optical system, and the resulting error is developed. Experimental data is then taken and found to agree with the mathematical prediction. From the mathematical relationship, an optimum structure layout is determined.

In lithographic metrology tasks such as overlay measurement, the pixel size of the imaging device is often much larger than the desired accuracy of the measurement tool. Interpolating the imaged data prior to the application of a measurement algorithm gives an accurate measurement but increases the amount of data to be processed, placing demands upon computational resources. In this paper, we propose a computationally-simple and highly- accurate algorithm for determining the subpixel shift between two waveforms. The algorithm utilizes the fast Fourier transform of each waveform to estimate the relative subpixel shift between the two waveforms using the complex phases of the frequency-domain-transformed data. The algorithm uses a simple iterative search for the position parameter that typically converges in less than ten iterations with optical data. We show that this algorithm estimates subpixel position more accurately than cross-correlation followed by interpolated-peak-fitting using both synthetic and real image data. Moreover, in high signal-to-noise ratio situations, the algorithm's precision can be shown to approach the fundamental precision limit determined by the statistics of the image data. Results of the algorithm's application to overlay data from a commercial scanning-confocal optical microscope are presented.

Systematic offset errors can be a major contributor to total overlay error on projection wafer steppers. To compensate for this offset error we have developed an automated method using measured overlay results from previous lots to generate offset correction values for current lots. This method provides statistically optimized offset corrections while eliminating the need to process test wafers. In addition the dynamic correction capability of the feedback control loop automatically compensates for offset changes such as drift in stepper alignment systems, process modification, and mask revisions. The system has been in use for over a year and has proven extremely effective in eliminating systematic offset errors on numerous devices.

For achieving more accurate overlay measurement of optical metrology system, this paper shows a new approach from the viewpoint of measured data correlation between before and after etching. From this comparison, two new methodologies are proposed. One method does not require measurement after etching to optimize measurement algorithms. The superiority of this approach is also discussed compared to the criteria of measurement repeatability and TIS. Using the new methodologies, measurement algorithms are optimized for many kinds of processed wafers, including high temperature sputtered aluminum, with a grainier surface than conventionally sputtered aluminum.

The wafer expansion scaling error in overlay accuracy using a stepper has been investigated. The scaling error depends on the exposure energy, mask aperture ratio and wafer surface reflectance. The scaling error is observed only along the X axis when using a global alignment method. From the experimental results, we present a mechanism which explains the scaling error. Then the occurrence of the scaling error according to the mechanism is simulated by the heat conduction and displacement analysis. The cause of the scaling error is proved to be local wafer expansion due to the accumulated heat generated by the exposure light energy.

This paper is a continuation of our earlier work which investigated means of reducing Tool Induced Shift (TIS) in overlay tools. In this paper we show that there can be significant errors in the measurements from a tool that shows zero TIS. These errors arise from proximity effects within the measurement target. Improving the resolution of the instrument reduces these errors, although this is counter to the current trend of the industry. We present the results of detailed comparisons of two measurement objectives. These have the same magnification but different numerical aperture so that they differ in resolution and depth of field. We will show that the lowest measurement error are achieved by acquiring all images from a common focal plane, even when the target marks are displaced vertically by more than 2 micrometers . Both objectives perform equally well, although we had expected that increasing the depth of field would provide benefits. The data shows that increasing the resolution of the microscope will allow the high accuracy and precision required for the future to be achieved without sacrificing an ability to measure the deepest structures likely to be met in these processes.

We discuss in this paper the use of the confocal microscope and an interferometric microscope for measurement of phase-shift masks. It is shown that phase measurements are more accurate than intensity measurements for determining the depth and slope of the sides of a notch. Results obtained on an isotropically etched mask, with curved bottom corners, are predicted well by our imaging theory. Solid immersion lens techniques for observing a mask from the flat side are demonstrated.

Amorphous silicon has been used extensively in electro-optical applications. Its use as a gate electrode material for advanced CMOS devices is currently being developed, as it offers certain desirable characteristics compared to the commonly used polycrystalline silicon. We have studied a series of amorphous silicon films deposited at varying temperatures. The deposition temperature ranged from 540 degree(s)C to 570 degree(s)C. The nominal thickness was approximately 1575 angstroms to 3170 angstroms. Due to the differences in deposition temperature, one would expect the optical properties to vary slightly. Using software that allows analysis of the spectral information, the dispersion was examined for each sample. With this knowledge, the film thicknesses could be reliably measured.

This paper examines the effects of specular and diffuse reflectivities, both empirically and theoretically, on the imaging process for highly reflective substrates. Aluminum (Al) and Titanium Nitride (TiN) wafers were used extensively in this study. TiN wafers were prepared with varying specular reflectivities and virtually no diffuse reflectivity. Dose to clear on TiN wafers was found to increase linearly with specular reflectivity in the absence of any diffuse component. Aluminum wafers were prepared with high specular reflectivity and varying diffuse reflectivities. On these Al wafers, the increase in diffuse reflectivity decreased the dose to clear in a nonlinear fashion. A theoretical explanation for the observed phenomenon is presented based on the interaction of thin film interference effects with diffuse scattering. Modelling results based on this theory are shown to be in good agreement with the experimentally observed data. Based on these results, a method for the generation of specification limits for the allowable variations in specular and diffuse reflectivities are also presented. Also discussed are the tools and methods used for measuring substrate reflectivity to obtain these specification limits.

A new self-reference signal processing technique is proposed for detecting the location of any nonregularities and defects in a periodic two-dimensional signal or image. Using high- resolution spectral estimation algorithms, the proposed technique first extracts the period and structure of repeated patterns from the image to sub-pixel resolution in both directions, and then produces a defect-free reference image for making comparison with the actual image. Since the technique acquires all its needed information from a single image, on the contrary to the existing methods, there is no need for a database image, a scaling procedure, or any a- priori knowledge about the repetition period of the pattern. Results of applying the proposed technique to real images from microlithography are presented.

Focused ion beam repair of opaque defects on 5X reticles cause a post repair stain due mainly to gallium ion implantation in the surface of the quartz substrate. The effects of these stains have been investigated through printability tests using a g-line optical lithography process. In some cases the effect of the post repair stain is worse than that of the original defect, highlighting the need for effective antistain processes for use in conjunction with FIB repair.

This paper presents an in-line monitoring scheme and zone partitioning experiment. An automated laser based wafer inspection system, linked with a defect data management station was integrated into the fab process flow as part of a defect reduction strategy. The results emphasize the feasibility of such a strategy as well as the ease of integration once such a strategy is established.

A high probability particle detection system for LSI wafers is proposed. In order to detect fine particles on bare wafers, optical noise from them are studied. Simulation with a diffraction model indicates that the optical noise is caused by diffracted light on the wafer roughness, and can be reduced by a large incident angle illumination and a detector with small pixel size. The side-scattering light detection system which has a illumination of incident angle of 80 degrees and a detector with a pixel size of 0.3 micrometers was confirmed experimentally to detect standard 38-nm particles in high signal-to-noise ratio, and the detection results were verified by SEM. Probability study of detection indicates that the detection probability reduces rapidly as detection light level is low. Two types of systems are proposed, a high detection probability system of 95% with energy of 600 mJ/cm² a low illumination energy system of 4 mJ/cm² with probability of 10%.

Printability of sub-micron 5X reticle defects at i-line and duv. wavelengths is assessed by use of a specifically designed test reticle incorporating defects whose size, and proximity to adjacent features, varies within sub-micron line/space arrays. Results are presented by plotting minimum resolved defect vs. array linewidth for both adjacent and isolated defect sites. Results on defect printability enable future reticle procurement specifications to be established. Results at i-line have been successfully modelled using SOLID which has a powerful graphics package enabling results to be displayed in 3D.

The spectroscopic film thickness measurement system is generally composed of a microscope, a spectrometer, and a data processing unit, and it enables non-contact measurement of transparent films (such as SiO₂) on semiconductor substrate with a minute spot objective of some micron meter in diameter. Its measurement principle is based on the interference effect of multiply reflected lights from film and substrate surfaces, and the measurement result is obtained by analyzing the spectral profile of the reflected lights separated by the spectrometer. The measurement accuracy depends on many factors. The accuracy has been already often discussed, however, few reports have referred in detail to the relation between the measurement accuracy and NA characteristic of objective. This report will deal with this NA-relating issue. This paper explains: (1) the factors that have an influence on the accuracy of the spectroscopic film thickness measurement, (2) influence by the NA characteristic of objective, (3) our spectroscopic film thickness measurement system, and (4) summary.

A new scheme for automatic defect classification and sizing is presented. The new scheme is developed for improving the overall production of automatic die-to-database reticle inspection equipment for defect detection. The new scheme replaces the time-consuming, inaccurate and non-repeatable traditional methods that are based on human reviewing and verification of defects with the aid of relatively crude image processing electronics. In order to overcome these limitations, a double-tier scheme has been developed for automatic defect classification and sizing (ADCS). The fundamentals of this scheme are presented. The image processing algorithms are described and their overall performance is evaluated using various test and production masks. The reported scheme represents a practical precise and accurate method for automatic classification and sizing.

Advanced semiconductor devices, such as 64 Mbit DItAMs, are characterized by increased circuit density, shrinking geometries (0.35 micron design rules) higher circuit complexity and an increasing number of mask levels used for device production. Successful manufacturing of such devices requires not only the decrease of the total number of defects, but also the decrease of the number of defects in each mask level. As a rule of thumb, defects which are larger than 1/4 of the mininum geometry may potentially cause device malfunction [1]. For devices such as 64 Mbit DRAMs this implies the need of monitoring very low defect densities of defects as small as 0.1 micron. These requirements give rise to new challenges and concepts in defect detection technologies. To meet in-line monitoring requirements, advanced inspection technologies should have: S Ability to cope with pattern densities and topographies which are below the resolution and depth of focus limitations. S Reliable detection of process induced pattern defects as well as micro-contaminations down to 0.1 microns. S High throughput. Present optical inspection technologies include CCD imaging [2], particle detection based on laser scattering techniques [3] and spatial filtering techniques [4], [5], [6]. An extensive survey of automated wafer inspection techniques can be found in [7]. Systems that utilize CCD imaging technique base their detection capability on high and distinct resolution of the pattern. This high resolution is essential for the detection of small differences in the images caused by the presence of defects. Since the smallest possible spot size is in the order of 0.6 micron, when dealing with 0.35 micron technology neither the pattern nor the pattern anomalies are resolved. In addition, defects which are smaller than the optical spot have inherent low contrast. All the above impose severe limitations on the use of CCD technique for micro-defect detection. Laser scattering techniques are used mainly for particle detection on patterned wafers. This technique is limited by its inherent poor resolution, or by pattern variations. For these reasons this technique is used only for moderate (0.5 micron and up) particle detection on layers after deposition as the detection capability is further deteriorated when scanning patterns after etch. Spatial filtering techniques use a blocking spatial filter located in the back-focal plane of the imaging optics in order to suppress the repetitive pattern signals, thus emphasizing the random defects signal. The two main drawbacks of this technique are the need for a specific spatial filter for each wafer type and its disability to inspect random logic devices. The authors believe that the above mentioned wafer inspection techniques have many shortcomings which prevent them from meeting wafer inspection requirements of the mid 90s. This stimulated the need for a new defect detection image acquisition concept. In this paper we describe a new wafer inspection technique which overcomes the limitations of presently available wafer inspection systems. The essence of this new technique is a detection sensor which utilizes a novel concept of Perspective Darkfield Imaging (PDI) combined with pixel-by-pixel die to die comparison. This provides for ultra fast and highly sensitive detection of micro-defects. A built in verification and classification capability enables the differentiation of particles from pattern defects.

Absolute pattern placement metrology on wafers yield significant improvements for tool design and tool engineering. Error limits needed for advanced lithography processes are presented, error sources involved in wafer metrology for characterizing steppers are shown. The algorithm used to extract the characteristic image of the reticle is outlined. Static repeatability of 3.3 nm for and of 8 nm for die grid was obtained (maximum, 3-standard deviations). Nominal accuracy was 7 nm (lens map) and 32 nm (die grid) as obtained by comparing measurements with wafer orientations of 0, 90, 180 and 270 degrees.

Abstract not available.

This paper presents a new feature extraction technique named the Bending Point Counting Method (BPCM) for detection of defects in microlithography patterns. Digitization leads to breaking up of continuous contours into line segments and adjoining bending points. The defects are detected by counting the number of bending points in a detector window having dimensions smaller than the minimum linewidth of the pattern. Due to applications of image processing techniques, the algorithm is able to cope with the image distortions and inaccuracies arising due to reproducibility of automatic setting of the inspection system. The paper deals with the principle and mathematical formulation of the algorithm. The experiments carried out to test the validity as well as versatility of the algorithm are discussed.

Technical requirements for advanced manufacturing processes require strict photoresist photospeed control. To meet this need, it is necessary to determine the error associated with measuring photospeed at both the photoresist supplier and user Fabs. This paper described two studies targeted at determining and improving the photospeed test error at both the supplier and user Fabs. The initial work focused on understanding each company's test methodologies, respective capabilities and sources of variation. This consisted of a 48-hour photospeed repeatability test during which process and environmental conditions were monitored. Through brainstorming and the results of the repeatability study, factors affecting the signal-to-noise were identified and evaluated in a designed experiment.

A new concept, the 'Application Specific Microscopy', is proposed here for today's complex lithography patterns containing 3D halfmicron information. The concept is based on the fact that a consistent 2D metrology process is no longer sufficient to characterize samples with 3D information. The Z information becomes necessary for the correlation between metrology measurement results to the sample XYZ topography. Also a complete Z information helps both metrology and lithography engineers to trace process variations. the Z information is in fact the FOCUS parameter. It becomes the most critical parameter in the XY halfmicron metrology process, similar to optical halfmicron lithography process. The Z information is acquired when the sample is scanned in Z. A Z-scan capability with nanometer resolution was used here to acquire the necessary Z information, an information that was then used to determine metrology system Depth Response Function. The Depth Response Function was used to monitor focus and to manipulate the metrology system for optimum performance on a particular sample.

A new approach to measuring submicron linewidths using optical phase-shifting interferometry is proposed. The technique has two main features. First, the surface topography is measured directly by determining phase information from the wavefront reflected from the surface of the object. Theoretical analysis indicates that the phase image of a line feature gives better lateral resolution than that of the intensity image of the same optical system. By comparing the intensity images and phase images from the computer simulations, it will be shown that the phase image (1) is affected less by the diffraction limitation; (2) has sharper edge definition; and (3) is insensitive to the variations of material composition and step height of the object. Second, the surface reflectivity can also be measured using phase-shifting interferometry. Unlike the intensity image measured by conventional methods, the measured reflectivity is not affected by any variations associated with the light source across the entire illumination field. The relative reflectivity between the line feature and the substrate is then determined. These advantages will result in better resolution and accuracy in measuring submicron linewidths. Theory and simulations predict that accurate measurements of 0.2 micrometers linewidths should be possible.

Linewidth control over reflective topography has become a major problem in sub-half micron optical lithography. Reflective notching and thin film interference are the major contributors to linewidth variation. In this paper, we examined a simple and practical approach to characterize photoresist swing curve and anti-reflecting coating materials using reflectance measurements. An optimization of titanium nitride (TiN) anti-reflecting coating (ARC) thickness for G-line, I- line and DUV wavelengths was also examined for aluminum substrates. Aluminum film reflectivity shows a correlation with surface roughness and reflectance measurement technique can be utilized to monitor thin films.

An aplanatic microlens is a small (typically 10 to 450 microns) diameter lens, which is placed close to the object being viewed and acts as a pre-magnifying lens for a microscope system. The lens is mounted on a stalk for easy positioning with a micromanipulator. All aberrations scale with size, so this small lens shows good chromatic correction. It is broad band even in the deep ultraviolet and even at high numerical aperture. The lens operates as a nearfield immersion lens when its planar surface touches the object being viewed. In this case the numerical aperture can be as large as n, the index of refraction of the material of the lens. For fused silica, a convenient material, this is about 1.46.

Precise measurement of pattern corner rounding is essential for determining the quality of current and next-generation photomask reticles. Using an appropriate test target with precise critical dimensions (CDs) and feature sizes, photomasks were written on an ETEC CORE 2564 laser writer with changes in focus from -1.0 to 1.0 micrometers at various exposure doses (mj/cm²). Using this methodology, the corner rounding of the CORE 2564 imaged feature was changed and compared to the ETEC MEBES III system generated pattern. The KLA 239HRS photomask inspection system was used to examine the effects of corner rounding at 0.25 micrometers with its most sensitive detectors at a 0.05 tolerance level. The optimization of writing tool performance and process parameters was enabled in two ways: by using the KLA 239HRS system, which did on-line biasing and rounding in 0.12 micrometers increments; and by using the Leitz AMS200 CD measurement tool, Micron Focused-Ion-Beam (FIB) system, and a binary image system, which measured the rounding of the mask features.

A new method is described for the real-time in-line control of critical dimensions for positive- tone chemically amplified resist systems. The technique relies on the generation of a diffraction grating in the resist film when a latent image appears during the post-exposure bake. A simple optical illumination/collection arrangement allows the diffracted signal to be measured during the post-exposure bake. This signal can be correlated to linewidths when measured by a non-destructive SEM. The result is a post-exposure bake time that can be used to correct for exposure-and-bake temperature variations to conveniently provide overall process control. Results generated by a prototype system are presented for a variety of 0.5- micrometers mask levels and process conditions.

Bright field optical microscopy has well-known nonlinearity problems due to varying film thickness. Phase images eliminate many of these problems. This paper will describe the advantages of phase imaging, based on calculated results. Experimental corroboration has been obtained with a bimodal coherence probe microscope. In the less coherent mode, the microscope utilizes illumination with a high numerical aperture and broad spectral bandwidth. In the coherent mode, the illumination is monochromatic and has a low numerical aperture. Experimental data will be presented and compared to theoretical results. The system demonstrates extended resolution for some materials and a reduced sensitivity to substrate variations.