Wafer stepper calibration and testing may be facilitated by using the auto wafer aligner feature of the wafer stepper as a metrology tool. This paper describes the procedures utilized in the calibration and acceptance testing of wafer steppers. The method of collecting data is explained, as well as the algorithms used to calculate calibration factors. Data from actual system tests is presented.

A direct means of observing both the location and the shape of the aerial image of a microlithographic stepper lens is described. Sub-micron resist structures, doped with a fluorescent dye, are swept through the aerial image as fluorescence is monitored. The resolution of this technique is not limited by optical wavelengths, but rather depends on the width of the fluorescent resist structures. Data are reported on the resolution and overlay error of a 5X reduction stepper. 60 hertz vibrational motions of the projected image with respect to the wafer were observed with peak to peak amplitudes as large as 0.3 microns.

We describe a new mask-to-wafer gap measurement technique, that has been implemented into the Electron-Beam Proximity Printing (EBP) lithography tool. In the EBP gap-measurement mode the parallel e-beam of 1 mm diameter performs scanning inclination about a perio-dic chip-registration pattern in the mask, resulting in the image of that mask pattern scanning across the wafer surface. Detection of strong scattering or reflection of electrons, when that image overlaps corresponding registration patterns on the wafer surface, reveals the image excursion in the wafer plane, which in turn is an accurate measure for the mask-wafer gap. An accuracy of + 1% for absolute mask-wafer gap data and - more important for application in EBP lithography - an accuracy of + 0.3% for relative mask-wafer gap variations was readily obtained over a gap range 560 μm to 1400 μm.

This article briefly reviews the unique features of the ASM Lithography wafer stepper that make it extremely well suited for precision metrology applications, and presents techniques and procedures developed for use in wafer-stepper manufacture (i.e., final assembly, final test, quality assurance, and field service). The procedures lead to unprecedented tight tolerances on some machine specifications, and to settings of maximum latitude for some other machine specifications. It is believed that these developments and other similar procedures will become requirements in machine manufacturing as the current trends in IC chip production continue to unfold (i.e., machine matching, tighter overlay design rules, larger exposure fields, larger wafers, etc.)

A new software option was developed for GCA wafer steppers which enables the site-by-site alignment system to be used as an automatic tool for stepper self-calibration and grid-matching. This software, called grid standardization, permits the use of ideal wafers to self-tune the stepper grid to a standard via site-by-site alignment prior to stepping product wafers. In this paper, the potential of this improved method of global alignment for use in VLSI circuit production is assessed by analyzing its various sources of misregistration. A mathematical model is developed which allows registration accuracy to be predicted for a given set of processing conditions.

In manufacturing integrated circuits, the tendency is toward smaller and smaller structures. Therefore the requirements for accurate micropositioning in lithography and wafer handling are growing more stringent. In many cases even a positioning accuracy of 0.1 micron is not good enough. But traditional precision mechanical positioners come to their limits in this range. Play, backlash, elasticity, friction and temperature changes can only be partially eliminated. One possible way to overcome these problems is with positioners based on the inverse piezo effect.

We report here the development of SMART SET', an expert system designed to complement the GCA DSW Wafer Stepperg, such that closed loop feedback for stepper set-up and calibration is realized. In particular, SMART SET' enables the user to monitor and optimize overlay and utilization performance. Measurements are obtained directly from the automated local alignment and history file systems of the stepper. An analysis package which embodies the expertise of stepper applications engineers provides rapid feedback of the system corrections required to sustain optimum stepper performance.

Unprecedented optical image resolution has been obtained using a near-field optical scanner called "optical stethoscope": 20 to 30 nm with blue light of half a micron wavelength! Key element is an extremely small aperture (5-10 nm) placed at the very top of a sharply pointed pyramidal or conical screen. The aperture is scanned in the immediate proximity of the surface to be investigated. The amount of light transmitted by both aperture and sample depends sensitively on the optical properties of the sample in the immediate vicinity of the aperture.

Theoretical models for optical imaging are used to investigate the performance of different linewidth measurement techniques. The harmonic content of the image profiles of semiconductor line objects is examined around the focal plane and it is concluded that focusing techniques which utilise the lower harmonics are unreliable. The analysis is used to determine the spatial frequency range for which image based focus detection is possible and the performance of suitable digital spatial filters is investigated. The sensitivity of thin and thick layer measurements to variations in parameters such as thickness, refractive index and wavelength is investigated and it is concluded that measurement variations between specimens may be reduced by using a broad band illumination source. A method of reducing the spread of measurements by contrast correction is presented. A linewidth measurement technique based on image scanning is discussed and typical repeatability performance figures are presented for a purpose built system.

We describe a new type of electronically-scanned confocal optical microscope which is very well suited to surface metrology. A laser beam incident on a wideband Te0₂ Bragg cell produces both a stationary beam which is used as a local phase reference and a scanned beam which is frequency-shifted; these are used in an AC interferometer. The system produces an RF output whose amplitude is that of the scanned beam, and whose phase is the optical phase difference of the two beams. Optical phase and amplitude can then be measured separately with standard RF techniques. Phase accuracy is about 10 . Transverse edge resolution (10%-90%) of 230 nm has been obtained (with λ = 514.5 nm) , which is in good agreement with the prediction of the simple theory. The data has been deconvolved to enhance the edge resolution, resulting in an ultimate resolution of 130 nm at the expense of introducing ripple into the edge response.

This paper will discuss the design criteria for an optical imaging system capable of high accuracy measurements on IC devices. In particular, the problem of measurement repeatability is discussed in the case of devices with submicron dimensions and complicated three dimensional features. Preliminary experimental results are presented from a confocal scanning laser imager designed for inspection and measurement on production wafers, SiScan-1 (TM). The system provides three major advantages over previously available optical microscopes: a) higher resolution and high magnification, b) accurate and repeatable C.D. measurements on micron and submicron geometries and, c) surface topography profiles and measurements. Digital images at high magnification (8,000X) and with accurately calibrated pixel sizes (0.06 microns) are illustrated. The problem of object edge to image edge correspondence is discussed. In particular, the profiling ability of the system will be used to illustrate potential solutions to this problem.

A laser induced photoresponse method is used for visualizing electronic properties of semiconductor materials and operational ICs. There are clear advantages compared to electron beam probing. By modulation of the laser light the method can be extended to internal logic analysis.

Single angle of incidence ellipsometrie measurements have been extended to dual angle measurements on our newly constructed multi-method precision ellipsometer in order to better determine the optical constants of a substrate. Following the measurement error analysis that was prescribed in an earlier paper for single angle of incidence and fixed wavelength measurements, the results for dual angle of incidence are presented here. Using an Explicit Error Analysis (EEA) method, involving the differentials of the measurable optical constants of the surface, it is possible to find a well-defined pair of incident angles to perform the measurement. Without a measurement error analysis, there would be no way of knowing what the absolute measurement uncertainty is or which angles of incidence could provide optimum measurement conditions. As in the case of single angle of incidence measurement where we were able to select an optimum angle of incidence to assure the highest measurement accuracy, the dual angle of incidence measurement also predicts optimum angles of incidence. It was found that in the case of single angle of incidence ellipsometry the principal angle of incidence can sharply define the optimum angle for measuring bare substrates and very thin films on a substrate. Likewise, for the dual angle of incidence measurement, there can also be two sharply defined angles for certain sample surface models. Here we present a dual angle ellipsometric measurement of the real part of the refractive index of a silicon substrate at the wavelength of 632.8 nm. A silicon dioxide film thickness between 125 and 150 nm and the two angles of incidence, 68 and 72 deg, optimized this measurement. The real part of the refractive index of the silicon substrate was found to be 3.865 ± 0.001.

Scanning tunneling microscopy (STM) is optimally suited to study surface topography and roughness at the nanometer scale where optical and other methods fail. The topography of nanocrystalline silicon is chosen as an example to demonstrate the outstanding potential of the new technique.

Current optical instrumentation being developed for critical dimension measurements in the integrated circuit industry is following one of two very different optical designs, i.e., either a focused laser beam which scans the wafer or the more conventional bright-field microscope. Traditional optical design lore has described these systems as "equivalent" based on the principle of reciprocity. More recent research has shown"- that the responses of these two types of systems are not equivalent for imaging of structures patterned in thin films such as those found in integrated circuit wafer fabrication. This lack of reciprocity is the result of the dependence of the diffraction pattern on the angle of incidence of the illumination. The impact of the lack of reciprocity on the design and calibration of critical dimension measurement systems is discussed.

The demands of submicron VHSIC Technology are trending to + .10 pm critical dimension control and + .10 μm overlay control. Equipment such as that evaluated in this paper will take an increasing role in this environment because it offers precision in the .002 μm range 1 a, moreover it produces the required data base in a reasonable time. The electrical prober provides alignment and linewidth data from specialized test structures etched into conductive films. This can provide a wealth of information about the photo/etch process. Again, it must be emphasized that the data collection process is highly automated and, therefore, very objective. In this paper, summaries of evaluations performed on various types of lithographic equipment available to us are presented. Processes used to generate the test structures are shown. Applications of the modeling capabilities when applied to the lithographic equipment are also included. This system can be very useful, for example, when applied to a projection 1:1 scanning system. One can model for translation and rotation error, commonly induced by operator misalignment, and thereby separate it from typical equipment related problems, i.e., scan, crosscan. The system will also model lens distortions and stepping errors from data accumulated with step-and-repeat equipment. Vector mapping of misregistration and contour mapping of linewidth variation gives one a powerful tool for seeing his own process capabilities. Comparison testing of the electrical prober to other linewidth measuring equipment, including optical and SEM, will also be reviewed. This is very important as linewidth measurement is fundamental to the electrical prober's operation. Optical systems for linewidth measurement and verniers placed directly on product provide a fine front line in baselining a process, particularly because of their immediate feedback of good data (bad sources notwithstanding). But, the process engineer using the electrical prober will realize a data base that truly enables him to make definitive statements about his own process as more stringent device requirements present themselves.

A whole wafer measurement method was developed to measure submicron oxide spacer dimensions. The technique offers the capability to evaluate the glass deposition and etchback processes used to form the spacers. Lateral diffusion of implanted layers can also be measured with this method. Measured spacer-width values agree well with measurements from SEM micrographs. Results show that the precision of this technique is presently limited by electrical probing accuracy and reticle fabrication.

The defect density of the metal layers in a double level metal process was studied. The dominant failure mode, with positive photoresist, was found to be intra-level metal shorts caused by particulate contamination. The major particle source was found to be deposited oxides used as dielectrics. Our processing environment, metal deposition process and lithography process were found to be relatively minor contributors.

Metrology techniques suitable for the characterization of X-ray lithography systems with submicron design rule capability are discussed in this paper. Registration and linewidth control are the main areas of investigation. Automated electrical measurements on special linewidth measuring test structures can be used for both registration and linewidth measurements. These techniques require some wafer processing steps. Other techniques that do not require special processing are the use of optical vernier structures for registration measurement and SEM analysis for linewidth measurement. Results obtained by these different techniques are presented and compared. Advantages, limitations and accuracy of each particular technique are discussed.

A parametric formula is presented which enables the determination of the CD control characteristics of a stepper from measurements of linewidth as a function of focus and exposure. The parameters obtained may be examined in terms of how they affect critical dimension control locally, and over the entire exposure field. Fundamental to the analysis is the ability to obtain large quantities of linewidth data. This problem is solved through the use of electrical test patterns. Experimental results of how this analysis has been used to examine several GCA steppers with state of the art reduction lenses is presented. A discussion is presented of how these resolution parameters are related to the reticle linewidth, exposure system, and various factors related both to the design of the stepper and processing of the wafers. Finally, an attempt is made to describe the exact physical meaning of each of the parameters in terms of their potential sources in the optical system.

An electrical probe technique is used to measure the widths of doped polysilicon structures. The measured values are compared to those obtained with a low-voltage SEM. This is done for widths in the micron to submicron range. A correlation is performed and various accuracies reported.

As the dimensions of semiconductor devices on wafers approach the micron range, it is readily apparent that measurement of these types of geometries on the fab line will not be obtained using optical methods (1-4). In addition, non-destructive scanning electron microscopy methods must be found to make more representative measurements than those provided by fracturing of the specimen or use of conductive coatings. Further, measurements of semiconductor geometries on wafers must be made in-process. That is, the measurements must be made on uncoated specimens so that the specimen can be returned to the process quickly and intact to add layers for both product and monitoring wafers. The primary objective of this paper is to demonstrate procedures for using Scanning Electron Microscopes for small feature measurements in Lithography. To this purpose, test wafers were generated containing resist valleys with varying widths and slope angles (resist features present the most challenging measurement objects). A selected number of these features were measured on Scanning Electron Microscopes. Recorded were SEM micrographs, backscattered electron profiles and numerical data. In addition, the same features were also measured on optical measurement microscopes and the discrepancies between different systems were analyzed. A second objective of this study was to assess the usefulness of continued optical measurements in lithography because they are faster and considerably less expensive and, therefore, should continue to play an important part in lithography. It was also found that (uncontrollable) resist slope and adhesion variations are responsible for a major portion of optical measurement errors. Finally, a method will be proposed for the application of SEM systems to advanced lithographic process and equipment characterization and monitoring.

This paper describes the use of a scanning electron microscope (SEM) for critical dimen-sion (CD) measurements on a semiconductor manufacturing line. The SEM was installed as a successor to optical measuring tools that were becoming inadequate for the task because diffraction and interference effects are limiting the accuracy of measurements, and resulting uncertainties are exceeding the reduced tolerances allowed in the production of electronic circuit patterns. During a year's time, the SEM system performance in processing some 6,000 wafer loads on a pilot line was investigated. Among the significant results of the investigation are the following: 1) Because the system resolution is limited to 200 A at 1 kV, comparisons must be made with measurements made on a high-voltage (25 kV) SEM in order to obtain proper edge parameter definitions. 2) It was possible to measure all mask levels with the SEM without sample surface charging hiding features. We also found no circuit damage traceable to the electron beam. 3) We found that it was possible to train operators adequately to achieve throughput rates (20-30 measurements per hour) which were comparable to those achieved with optical tools. 4) By observing human factors in the use of this complex system, we were able to improve ease of learning and ease of use by line operators, increasing the system's acceptability to these people, thus assuring better measurements in the manufacturing environment.

In microelectronic applications, electron beam measurements are frequently plagued by electromagnetic interference and vibration in the fabrication environment. In the past, efforts to control environmental effects have focused on attempts to decouple the measurement apparatus from the source of interference. Rerouting power cables, balancing phase loads, shielding with mu-metal and copper screen, isolating on concrete pads, and mounting on vibrationally damped columns have all been tried with some degree of success. The com-mon problem with these active solutions is that retrofitting an existing facility can be prohibitively expensive. We have found, however, that where environmental effects are relatively well known, they can be removed from the data a posteriori. In particular, 60-Hz electromagnetic interference, appearing as sawtooth patterns in non-synchronous SEM images, is probably the most common deleterious effect experienced in fabrication areas. Techniques are demonstrated by which tuned integrating and differentiating filters can automatically reduce the measurement error by as much as 80 percent.

The National Bureau of Standards is currently developing a new scanning electron microscope-based linewidth measurement system for future calibration of standard reference materials for the IC industry. This system incorporates a piezo/interferometric stage for precise translational motion and the monitoring of distance, improved vibration-isolation, microprocessor stage control system, and computer data analysis. The specifications incorporated into the system are designed for the measurement of linewidth dimensions from 0.1 to 2 μm with a precision of 0.002 μm. The design philosophy of the system is discussed along with the current limitations of accurate edge detection in SEM-based systems.

As the latest integrated circuit designs move toward submicron feature dimensions, optical microscopes commmonly used for critical dimension (CD) measurements no longer have sufficient resolution. 1,2 Just as the semiconductor industry has turned to electron beam lithography for mask making and direct writing to achieve submicron dimensions in VLSI and VHSIC devices, so too must it turn to electron beam instruments to measure these devices. A computer based system for digitally controlling a scanning electron microscope (SEM) is described and its ability to provide precise, accurate measurements evaluated.

Minimum linewidths of integrated circuits in mass production are currently (1985) just above lum. They are projected to decrease to 0.7um and then to 0.5um in the next decade. Linewidths currently must be controlled to ±10% and it is predicted that linewidth control may need to be ±5% in the future. To reach present goals, linewidth measurement systems with accuracies of at least 10% of linewidth (3 sigma) and, in some cases, ±5% (3 sigma) are required. Errors less than 0.05 to 0.lum (3 sigma) on lum lines and 0.025 to 0.05μm (3 sigma) on 0.5μm lines are therefore needed. This precision must be maintained over long time periods, with different operators and when thicknesses, profiles and other parameters of the structures change due to normal process variations.

We have developed the Model S-6000 for critical dimension measurement of circuit patterns on VLSI wafers utilizing a filed emission electron source. This instrument has been designed to handle up to a 6" wafer. Measurements are made under low voltage operation, at a high resolving power and at a flicker-free TV-scan rate. Minimum dose feature protects the wafer under measurement from electron beam damage. The S-6000 is a state-of-the-art CD measurement, SEM for quality control of in-process wafers.

Miniaturization in many industries has generated a need for sub-micron measurements. The most significant need is in the semiconductor industry for reproducible integrated circuit critical dimension measurements essential to characterizing circuit geometries. Line width measurements to date have been performed with light microscopes. As circuit element dimensions shrank toward the sub-micron region the fundamental resolution of the light microscope became the limiting factor. Conventional SEMs, with far greater resolution and depth of focus capabilities, were met with disappointment because they were designed as multi-purpose instruments without consideration for the unique problems associated with in-process SEM analysis of large, poorly conducting wafers in an environment of high magnetic fields and vibrations. The SCANLINE has been designed from the floor up to address these problems and to be an easy to operate sub-micron production measurement tool.

VLSI and ULSI developments aimed at sub-micron geometries require improved measuring systems. A computer controlled cassette to cassette electron beam system has been developed. High resolution of l5nm at lkV is routinely achieved with a unique 4 lens electron optical system. Through-put of the system is 10-15 wafers per hour (5 chips, 1 measurement/chip). An automatic line width measurement method with high accuracy, "The Linear Regression Method"1, has achieved an absolute accuracy of 0.009μm.