As high-volume chip production presses into submicron design rules, parasitic error sources in imaging and overlay may be reduced still another factor of two. Variable magnification, and magnification and focus corrections to compensate for barometric pressure changes, for example, illustrate how some imaging/overlay error sources can be dramatically reduced. Another error source in imaging/overlay now receiving considerable attention is lens distortion; lens distortion on the order of 0.2 micron can be a substantial fraction of the error budget for overlay between matched wafer steppers in production of chips with submicron design rules.

This paper establishes the characterization techniques that are needed to repeatedly fabricate submicron structures which are below the resolution specifications of an optical system. A two-parameter system is defined, consisting of a material or photosensitive component and an optical component, which represents the major contributors to process variability. Statistical inference and correlation analysis are used to estimate the effects and extent of the individual components on the total lithographic system. The photosensitive component, consisting of the photoresist, films and related chemistry, is shown to substantially represent the variability in the final linewidth of 0.5 μm structures. Positive and negative photoresist systems are compared and analyzed for run-to-run and within-run variability utilizing photoresist speed-point analysis. The optical component is explained by using a two-dimensional aerial image model which takes as input the numerical aperture, wavelength, coherence, aberrations and experimentally measured effects. A method for quantitatively measuring flare or background exposure is introduced.

Depth of focus requirements and contributions to the focal error budget for submicron optical lithography are reviewed. Models are presented which estimate depth of focus in both thin and thick layers of photoresist. The effects of resist refraction on usable depth of focus are considered. Measurements of image plane tilt, curvature, and astigmatism in 5X reduction lenses collected using an automated, in situ, aerial image monitor are analyzed.

The proximity effects are becoming a significant consideration to the linewidth control in sub-micron optical lithography, as the lithographic tools are being operated closer to the resolution limit. There are various forms of proximity effects. In this paper, we will attempt to classify the observed proximity effects in optical lithography and analyze their potential impact to device fabrication. The results of SAMPLE simulation and experimental characterization will be presented. Our conclusion from this study is that a method of linewidth compensation must be developed in order to achieve optimum device density and yield, using sub-micron optical lithography.

New lenses for i-line lithography will be discussed. The advantages of shorter wavelengths for achieving higher resolution with greater depth of focus are well known. While prototype i-line (365nm wavelength) lenses have been available for several years, the field sizes were too small to be useful in production. Several lenses have become available recently which offer higher resolution and/or better field size. Among these are the Tropel 1635i and 2235i lenses, 5X lenses with numerical aperture of 0.35 and resolution of 0.8um. Field sizes are 16 and 22mm. Experimental data on resolution, depth of focus, CD control and image placement errors will be presented. In addition, several i-line resist processes will be discussed from the standpoint of optimum resolution with single layer processing.

In this paper, the image resolution and depth of focus of a 10X 0.42NA i-line wafer stepper is shown. The effects of photoresist processing on these parameters are demonstrated. The processing materials used include i-line resists and contrast enhancement material. Additionally, a machine approach to dealing with the limited depth of focus of this high numerical aperture lens stepper is investigated.

Optical steppers have opened their new era beyond 1 Am barrier. In order to achieve submicron lithography, however, steppers have been obliged to solve many problems that were not noticed before. For example, reduction lenses that were recognized as stable components have now become dynamic and intelligent system. There are three basic parameters in the projection optics---- resolution, depth of focus and distortion. As for resolution, the uniformity of the image quality over the whole field is required. The SEM image of 0.8 Am and its uniformity are discussed and some simulation results are shown. Depth of focus is the second point and is recognized as one of the most critical parameters in optical lithography. In order to achieve large depth of focus, new optical auto-focus system is introduced. This system has two features. One is the grazing angle of incidence and the other is the multiple wavelength effect. The third point is distortion, which is the final key for overlay accuracy. The stability of distortion due. to barometric compensation and the machine to machine mixture problems will be discussed in detail.

Higher numerical aperture and shorter exposure wave length are important in attaining high resolution step and repeat photolithography. However, the usable depth of focus tends to be shallower when using higher numerical aperture lenses and shorter exposure wave length. This contradiction is a major obstacle in submicron photo-lithography. In this paper, a novel method for enlarging the defocus tolerance, Focus Latitude enhancement EXposure method (FLEX) is presented. FLEX is a method that focuses the image on the upper and lower levels of the steps on the wafer surface and takes multiple exposures of the same position under different focusing conditions. This method was applied to the i-line(365nm) stepper with 0.42 numerical aperture. A practical focal range expanded to 5μm. This amount of defocus tolerance is approximately three times larger than that of the conventional method. In FLEX, the image contrast of line/space patterns are lower than that of the conventional method. However, the contrast enhanced layer improves the image contrast greatly. Using FLEX, submicron contact hole patterns are successfully delineated over a 10μm step height.

Numerous advancements based on the inherent simplicity and excellent optical correction of 1:1 broadband stepper lenses have occurred over the past year. This paper briefly touches on the lithographic requirements of various segments of the semiconductor industry and describes photolithographic developments that meet many of these requirements. Comparisons will be made between currently available 1:1 lenses with fields large enough to contain multiple four megabit DRAMs and a new lens with a resolution and field size sufficient to produce 16 megabit DRAMs. This new lens has a variable numerical aperture that allows it to achieve a minimum feature versus maximum field size tradeoff thereby accommodating different industry requirements. Important process parameters relating to process optimization and focus latitude for submicron lithography are discussed. Recently applied technologies such as voting lithography, that may be useful for device prototyping, are presented. Additionally, topics such as multiple field reticles for ASIC applications, improvements in 1X reticle technology, and improvements in overlay that will allow the next generation of DRAMs to continue to be produced in a multiple machine environment will also be discussed.

Full-field (e.g., steppers) and ring-field (e.g., scanners) systems can be expected to differ in resolution, depth of focus, and exposure latitude as a result of their optical and opto-mechanical designs. In this study we compare the results of performance simulation of two exploratory optical systems, both having the same nominal resolution. Working values of resolution and depth of focus are determined based on the approach of B.J. Lin. The impact of wafer unflatness on the focus budget is assessed by measurement of a batch of six-inch wafers. Various focusing algorithms are simulated, leading to a measure of the reduction in focal tolerance due to wafer unflatness for both ring- and full-field systems.

Integrated circuit (IC) design rules are determined by: 1. Circuit performance required to meet the needs of the IC customer, 2. Production capability of the IC manufacturer. In the first case, the only constraint is device physics. In the second case, the constraints are equipment and process performance coupled with the intrinsic variability of the manufacturing cycle. We view the generation of design rules from the manufacturing perspective. In examples, we relate design rules to lithographic performance as measured in a one-micron production environment. The effects of spatial and temporal variation on critical dimension and pattern registration control are treated. We establish a method to identify the cost/benefit tradeoffs of design and manufacturing. The resultant design rules embody guidelines for the cost-effective manufacture of competitive IC products.

An evaluation of a wafer stepper with the new improved Philips/ASM-L phase grating alignment system is reported. It is shown that an accurate alignment system needs an accurate X-Y-0 wafer stage and an accurate reticle Z stage to realize optimum overlay accuracy. This follows from a discussion of the overlay budget and an alignment procedure model. The accurate wafer stage permits high overlay accuracy using global alignment only, thus eliminating the throughput penalty of align-by-field schemes. The accurate reticle Z stage enables an intra-die magnification control with respect to the wafer scale. Various overlay data are reported, which have been measured with the automatic metrology program of the stepper. It is demonstrated that the new dual alignment system (with the external spatial filter) has improved the ability to align to weakly reflecting layers. The results are supported by a Fourier analysis of the alignment signal. Resolution data are given for the PAS 2500 projection lenses, which show that the high overlay accuracy of the system is properly matched with submicron linewidth control. The results of a recently introduced 20mm i-line lens with a numerical aperture of 0.4 (Zeiss 10-78-58) are included.

One of the most important -features of photolithographic equipment is the automatic wafer alignment capability. As the resolution of wafer steppers approaches 0.5 micron, alignment accuracy must be improved correspondingly. In addition, the wafer alignment accuracy must be insensitive to process variations. However, recent photoresist technology developments have introduced a number of highly absorbing coatings for the sake of improved resolution and better linewidth control. This includes the use of heavily dyed photoresists, antireflective coating layers, and contrast enhancement materials. With such highly absorbing photoresist coatings, it becomes difficult to see the alignment target under these coating layers. In order to achieve alignment, it is often necessary to pre-expose and develop the resist over the target, or to optically bleach the photoresist over the target by pre-alignment exposure, or to blast away the photoresist over the target using excimer lasers. In the mean time, a much better and much easier to operate approach using non-actinic alignment wavelength microscope 'b has been developed by ASET. This new approach overcomes the difficulty of seeing the alignment target by shifting the alignment wavelength to the blue-green part of the optical spectrum where practically all of these exotic process layers become transparent. It is also compatible with existing TV based alignment microscope regardless of whether it is set-up with the bright field or the dark-field schemes. In this paper more technical details of this subsystem will be reported.Data acquired on some of the most difficult customer wafers will be presented.

The successful implementation of advanced lithographic systems into high volume semiconductor fabrication has been hampered by the inability to monitor overlay using actual production wafers. While electrical probe techniques provide the necessary precision to characterize a lithographic system, special wafers and masks or reticles are required. This paper describes a new overlay measurement system (hereafter referred to as OMS) developed to use small targets placed on actual production wafers and masks or reticles. Using a simple detection scheme and a standard microscope, the system can rapidly measure up to 1000 sites per wafer. The system is designed for a precision of better than 0.05 microns (3σ) and measurement time between 1.5 and 3 seconds (including travel) per site. The system software can present vector plots as well as individual error coefficients. Resultant distortion, alignment, and magnification errors are calculated to aid in the optimization of the stepper or scanner system overlay.

A series of models are presented which enable optical alignment signals to be computed for a wide range of target structures. Scalar diffraction theory is used to describe the illumination and imaging processes and a waveguide model is used to describe the way in which the alignment target scatters light. A generalised imaging configuration is used to model four alignment systems; bright field, scanned dark field, key and target convolution and interference of the first diffraction orders. The effects on alignment accuracy of asymmetric resist flow over the target are studied as well as the effects of focus error and comatic aberration. It is concluded that bright field alignment systems are unlikely to provide the alignment accuracy required by micron and sub-micron lithography.

A fully automated direct reticle reference alignment system for use in step and repeat camera systems is described. The technique, first outlined by Janus S. Wilczynski, ("Optical Step and Repeat Camera with Dark Field Alignment", J. Vac. Technol., 16(6), Nov./Dec. 1979), has been implemented on GCA Corporation's DSW Wafer Stepper®. Results from various process levels covering the typical CMOS process have shown that better than ±0.2μm alignment accuracy can be obtained with minimal process sensitivity. The technique employs fixed illumination and microscope optics to achieve excellent registration stability and maintenance-free operation. Latent image techniques can be exploited for intra-field, grid and focus characterization.

ontrol of wafer temperature is essential for resist survival in plasma etching processes. Real time measurement in the reactor is difficult using conventional methods. In this work, a non-contact measurement technique is employed to study actual wafer temperatures as a function of process parameters. Various methods of thermal coupling between the wafer and the electrode are compared.

Two dry etch technologies are compared for application in dry development of resist. All results are obtained using the DESIRE [1,2] process. We show that MIE and RIE can give good results. Experiments were performed using a planar reactor and a hexode reactor in reactive ion etching mode. Although the develop rate is about ten times slower in the hexode reactor, selectivity and image quality is comparable to the planar reactor. Using magnetron assisted etching very high etchrates up to 3 μm/min can be obtained, but the etchrate is flowrate limited. The selectivity can be higher than in the pure RIE set-up but optimisition of the image does not require very high selectivity. The effect of the develop conditions on the lithographic contrast of the resist system are discussed. The contrast can be varied between 2 and 5 by changing these conditions.

A new trilayer sandwich structure has been developed to achieve tapered contact sidewalls and improved metal step coverage. Following furnace reflow of the conventional bilayer sandwich, undoped oxide is deposited across the wafer. When contact holes are wet etched through this trilayer sandwich, the sidewalls become well tapered with significant improvements in metal step coverage throughout the contact cuts. Incorporation of an additional undensified oxide layer over the existing bilayer sandwich has reduced the etch rate at the oxide/photoresist interface, improving metal step coverage and extending the utility of our wet etch without any change in etchant bath concentration or temperature.

It has been a long-standing goal of lithography engineers to match their exposure systems, using an independent monitor, such that a single exposure setting would deliver an identical effect to photoresist from any of their machines at any time. This is a relatively trivial issue for monochromatic systems, since both the photoresist and the detector system are unaffected by spectral sensitivities. However, the problem becomes more complex in the case of polychromatic systems such as the Perkin-Elmer aligner or the Ultratech Stepper. This paper outlines some of the considerations in pursuing this goal, and describes the implementation of such a system in a manufacturing facility utilizing polychromatic wafer stepper systems.

The trend toward tighter design rules and larger chip sizes have driven the demand to obtain the ultimate performance from present steppers. In this study, the G-line steppers with 0.30 NA lens are used to print 1 micron geometries. It is found that the contrast enhanced resist process is feasible to print 1 micron geometries and a single layer resist process can be utilized to print 1 micron square contact by applying a sufficiently thin layer of resist to obtain a conformal coating over the topography. In addition, big contact on the mask should be sized to smaller sizes so that over-exposure can be utilized to open small contacts.

Liquid particle counting (LPC) based upon laser light scattering technology continues to find wide applications in semiconductor manufacturing. Because defect levels in semiconductor devices are related to cleanliness of the processing chemicals, both users and vendors of photoresists and auxiliaries have expended efforts toward LPC testing programs. This paper describes Dynachem's approach to solving problems that have made vendor-user correlations difficult and have prevented setting of meaningful specifications. The paper also discusses how statistical process control can be applied to production of photoresists and developers. As demonstrated in the following discussions, the combination of statistical process control and particulate analysis is powerful and should help vendors and users move toward their goals of specification development.

A method for computing the images of contoured layered media is presented. The wave equation is solved in the layers by invoking the Rayleigh hypothesis, and the solution is cast in the form of a scattering matrix which relates the Fourier components of the incoming field to those of the outgoing field. Images of layered structures produced by a given optical system can be calculated by combining the scattering matrix representing the object with matrices representing the illumination and imaging optics which include the effects of apertures and aberrations. Results from the application of this technique to the problem of imaging alignment marks under photoresist will be presented and compared with experiment.

Basic studies of projection printed images are presented to identify the types of patterns which are most susceptible to residual aberrations and to establish test structures which may be used to monitor the presence of critical types of residuals. These effects are explored by including arbitrary lens optical path difference (OPD) aberration functions in a two-dimensional optical image simulation program associated with SAMPLE. The lens aberration function is expressed either in Zernike polynomials or a series expansion. The intensity is calculated from Hopkin's transmission cross-coefficient formula-tion with a self-checking algorithm. A catalog of results is presented here for the dominant primary aberrations of coma and astigmatism for a fixed maximum OPD of 0.4 λ. Contact holes are shown to be much more susceptible to astigmatism than coma and the traditional checkerboard test pattern is verified as a sensitive diagnostic pattern. An alternative structure con-sisting of thin lines with a short break is shown to be even more sensitive to astigmatism and useful for distinguishing it from coma. A further improvement in sensitivity is obtained through the use of small nonprintable defect-like features in proximity to regular features which coherently interact with the blurred image of the feature. A test target of this type is recommended for monitoring coma.

Overlay, throughput and lens performance are three critical parameters of optical alignment equipment. High overlay accuracy will give high die yields for alignment sensitive parts or can allow designers to use tighter design rules which will allow a shrink of the chip and more chips per wafer. This work contains high overlay accuracy data obtained during the processing of production wafer lots of a 64K HRAM (Hierarchical RAM) device using an ASM Lithography PAS 2000 Wafer Stepper.

A prototype automated photobay was installed in an existing fab area utilizing flexible material handling techniques within a clean tunnel. The project objective was to prove design concepts of automated cassette-to-cassette handling within a clean tunnel that isolated operators from the wafers being processed. Material handling was by monorail track transport system to feed cassettes to pick and place robots. The robots loaded and unloaded cassettes of wafers to each of the various pieces of process equipment. The material handling algorithms, recipe downloading and statistical process control functions were all performed by custom software on the photobay cell controller.

The physical concepts and yield advantages of pellicle protected masks have been reported in many papers. Pellicle disadvantages such as fragility and the hazard of particles trapped under the membrane have been greatly outweighed by the observed chip yield increase. Yet the pellicle provides no protection against static discharge, the dislodgement of previously electrostatically fixed particles, pellicle degradation and surface etch faults.

Local wafer thickness variation is a significant source of defocus error for submicron lithography. A method of characterizing the flatness of wafers using the auto-focus system of a stepper is developed. An analysis of the data is carried out by a program that provides the distributions of TTV (total thickness variation), LTV (local thickness variation) and LAV (local averaged thickness variation). Contour mapping of the average height per field across a wafer can reveal the areas where the most or least rapid thickness changes are occurring, which can be caused by vacuum bending, chuck tilt, or the different polishing techniques used by vendors. The areas of investigation include a comparison of samples supplied by several silicon vendors, the effects of edge exclusion, and the impact of typical MOS process steps. Wafer flatness is correlated to critical dimension control. In the worst case, local thickness variation within a typical stepper field of more than four microns is observed on new, silicon wafers which causes resist bridging over a large portion of an image field. Finally, the potential benefit of wafer leveling systems is discussed.

Contact hole array imaging is compared to line-space grating imaging to determine the resolution of contact hole features printed by stepper lenses. Both calculated and experimental results are reported. Image intensity profiles, IIPs, have been calculated for selected cross-sections of square contact holes and equal line-space gratings. The calculations are done according to the method of Hopkins for circular aperture optical systems, partial coherence levels, fractional pupil fill (S), of 0.5 and 0.7, and various levels of optical defocus. The calculated results show that for S = 0.7, at resolution levels (i.e., half-pitch of equal line-space grating or the width of the square contact) greater than 0.9 λ/NA, the IIP of the lateral cross-sections of contact hole arrays to be equal to or better than that of the correspondingly dimensioned gratings. At resolution levels less than 0.9 λ/NA, the contact hole IIP degrades rapidly with respect to the grating IIP. A comparison of the lateral and diagonal cross-sections of contact holes with no defocus aberration shows that the shape of the contact changes from a 'square' to a 'circle' in the 1.25 to 1.00 λ/NA range. In the 1.0 λ.NA resolution region the lateral contact hole IIP has a greater intensity than that of the space of the grating. The effects of optical defocus on contact hole imaging are examined for g, h, and i-line stepper lens at the same λ/NA resolution levels. Contact hole IIPs degrade more rapidly with optical defocus than the correspondingly dimensioned grating IIPs. Experimental comparisons between contact hole and grating resist images are presented for several stepper lens systems. The comparison between calculated and experimental results shows significantly less than 'perfect' imaging for some stepper lenses. Contact hole imaging characteristics are much more sensitive to residual aberrations in the optical system and can be used as a relative measure of the magnitude of these aberrations.

Wafer steppers play an important role in VLSI production. However, wafer exposure by a stepper requires both a perfect or defect-free reticle and an elaborate resist process. When a reticle is produced by mask shop its perfection or pattern integrity is guarantteed with a reticle inspection system. When introduced and used in the VLSI production, however, the reticle sometimes becomes defective due to falling particles and/or chrominum peel-off caused by inadequate reticle cleaning. Even with a pellicle, the reticle gets contaminated with dusts caused by careless handling. Therefore, the in-situ reticle inspection or reticle qualification is necessary to secure the exposure of product wafers.

Die-to-database inspection allows a repetitive or non-repetitive pattern of any scale to be examined for data errors, random defects, and repetitive defects. This type of inspection finds repetitive defects that would not be detected in a die-to-die inspection mode. Increasing circuit and process complexity requires more intricate methods of data handling in terms of layout, sizing, and merging from design to final pattern generation. This increase in complexity of both data and data manipulation enhances the probability that an inadvertent data change may occur at some point between layout and final product. Inspection of the mask or reticle to the verified design will either find these unwanted changes or verify that no changes have occurred. Use of the Calma database as the source for the inspection provides a separate data conversion path parallel to the path used for pattern generation. This allows a more extensive data verification to be performed. If an intermediate format such as Electromask, Mann, or Mebes is used as source, some types of errors would not be detected since the data used for pattern generation and inspection would be one and the same. Thus, both the final product and the methods of data manipulation needed to produce the final product are checked when the Calma database is used as the inspection source. Pattern defects found with this type of inspection will be presented and explained as to their cause. The factors and limitations inherent to pure database inspection using the KLA-221 Klaris inspection system will be discussed.

ATEQ Corporation has developed a high precision, high performance optical lithography system, the CORE-2000, based on scanning laser technology and optimized for the high speed writing of 5X and 10X reticles. A detailed evaluation of this system has been made using Nikon 21 measurement systems, KLA inspection systems, and throughput benchmark test patterns. This paper briefly describes the CORE-2000 and presents the results of this evaluation including registration performance, critical dimension control, defect control and throughput.

Glass wafers and automatic defect inspection equipment are being used successfully to do reticle image qualification. This technique assures that a reticle is free of repeating defects prior to being used to expose silicon wafers on a wafer stepper. The task becomes more essential, and more difficult as geometry sizes break the 1 micron barrier and relevant defect sizes approach 0.35 micron. In this study, the printibality of defects on glass wafers is examined and defect size is compared to the same defects printed in resist on silicon. Glass wafers are then inspected on a KLA 209. The concepts of defect fidelity function and effective defect capture size are defined and experimentally validated for wet and dry etched chrome and aluminum clad glass wafers. Undercutting is also examined for the four cases.