In the vast world of integrated circuits mask making is often taken for granted. This was particularly true a decade ago when the availability of a commercial a-beam machine, MEBES, considerably improved the accuracy of photomasks and simplified the manufacturing process. At present we have the capability to meet today's needs [~ 0.9 micron design rules] but we do not have the capabilities for the next reduction in design rules [~ 0.5 micron in ~1990/1]. Pattern generators, resist, measuring equipment, and defect detection are all suspect as we push photomask tolerances into the sub-micron region. This talk will review some of the limitations in today's photomask fabrication and some of the opportunuities that lie ahead.

The photolithographic properties of scanned laser imaging systems are discussed in comparison with conventional optical projection printers and e-beam imaging systems. Because a scanned laser system can effectively stitch together a large number of small fields, the customary trade-off between resolution and field size does not have to be made, and higher NA lenses can be used. Residual system astigmatism can be corrected by adjustments to the input illumination of the lens. The imaging properties of a Gaussian beam system are fundamentally incoherent and are a strong function of the input beam truncation. Contrast curves are developed for different amounts of truncation and compared with the theoretical and experimental contrast of projection printers. Although e-beam systems can resolve smaller features than optical systems, the effective image contrast of features substantially larger than their minimum spot sizes is reduced by electron back-scattering. E-beam contrast curves are compared with those of modern optical lithography tools. Finally, the resolution limit for laser scanned lithography is explored based on available deep UV CW laser light sources and high NA lens technology: it is estimated to be below 0.2 μm.

Improved resolution of an available i-line (365nm) stepper using a phase-shifting mask is discussed. The resolution investigated here is not only for periodic lines but also for isolated spaces and hole patterns. To obtain a narrow bright line for printing a fine isolated space on a wafer, two additional line apertures with widths smaller than the critical dimension of the stepper lens are placed on each side of the main aperture of the mask. The optical phase of the main aperture and those of additional apertures are opposite. The additional apertures play a role in reducing the bright feature size to less than the line spread function of the lens. Similarly, printing a fine hole is accomplished by using a main aperture surrounded by four additional apertures. The intensity distribution on the wafer is calculated by comparing the results obtained with a phase-shifting mask and those obtained with a conventional transmission mask. Patterns are also printed on the wafer using an i-line stepper with a nominal 0.55 μm resolution. A pattern of 0.3-μm lines and spaces, 0.3-μm isolated spaces and 0.4-μm hole patterns are resolved using the phase-shifting mask. This resolution is impossible with a conventional transmission mask. The effects of variations in the optical phase of the additional apertures are also investigated. The intensity calculations and experimental results suggest that it is possible to control the position of the best focal plane by changing the optical phases of the additional apertures.

Application of serifs in microlithography helps to extend and maintain more aggressive groundrules. Furthermore, this method of mask correction can be implemented in a conventional mask making environment. It puts no new demands on lithographic printing tools, and helps to extend their useful lifetime. Images whose dimensions are near the resolution limit of the imaging system are recorded in photoresist with size and shape different than predicted by geometrical optics. This difference is referred to as print bias. For small 2-dimensional features it can be characterized as shrinking of lateral dimensions and a corner rounding effect. It results in the loss of area, perimeter, and aspect ratio of printed features and makes it difficult to simultaneously maintain a desired tolerance for all features in the design. Modelling of optical projection printers is used to characterize printability of rectangles in a wide range of sizes and aspect ratios, with printed area being the primary parameter of concern. Motivation for use of the serif-like correction of reticle shapes is examined and a simple design procedure using experimental data on printability of small squares is developed. This procedure yields a serif design which, when added to reticle features, results in the area printed more closely to the nominal value and provides an optimal compromise between the compensation of the corner rounding effect and the associated process variability. A single size square serif is shown to correct a broad range of shapes and sizes. Improvements in terms of reduced radius of corner rounding, reduced area loss, and better preservation of aspect ratio of rectangles are shown. Use of serifs is also shown to result in a larger range of focus conditions for which, given a set exposure variation, all design features print with defined tolerances, i.e., a larger common process window. SEM microphotographs of resist images illustrate the use of serifs to improve lithography of small rectangles. Experiment confirms a substantial reduction of corner rounding and compensation for the loss of area, and aspect ratio in rectangles. An example of application to a dense pattern with typical circuit characteristics is also given.

Due to the ongoing development of submicron devices for mass production, the reticles used in the masters for forming patterns are being subjected to increasingly stricter quality requirements. For 5X reticles, even a minute defect less than 0.5um in size is transferred to the wafer. Moreover, a wide variety of defects must be detected and eliminated. There include complex pattern layout errors, substrate defects, and transparent foreign substance, as well as simple white or clear defects. Regarding the sizes of defects, recent improvements in defect inspection machine have made it possible to detect defects as small as 0.2um, yet it is evident that existing inspection technology is not able to cope with an array of new defects that were previouly not detected during the pre-submicron process.

To determine the effect of wafer stepper resolution on the printability of reticle defects, a test reticle containing edge-related submicron programmed defects was printed using 0.35, 0.43, 0.45 and 0.54 NA g-line 5x steppers. The size of the minimum printable defect was determined by examining the wafer photoresist images using both optical and scanning electron microscopes. Photographs show the effect of submicron defects located on or near geometry edges. Submicron reticle defects affected critical dimensions, image profiles and resist thickness in a 1.5 λ/ NA diameter "sphere of influence" on the wafer. The higher resolution, higher NA steppers printed smaller defects. Defocus was observed to increase the damage caused by printed defects.

A method for repairing transparent defects in photomasks has been developed. We discuss the experimental development and integration of the process into a commercial repair system. The defect repair process makes use of a laser to initiate thermal decomposition of a mod-ified commercially available gold metallorganic ink. The ink is locally decomposed with a laser over the defect thereby forming an adherent opaque gold repair pad over the defect. Rectangular repair pads from lum to 25um on a side can be deposited with a single laser exposure of 5 seconds.

Geometrical patterns with high sensitivity to optical imaging system parameters have been incorporated into electrical test masks and evaluated. The exploratory structures consist of targets such as fine-lines and small-dots, checkerboards with modified sizing, subimageable defects which interact with features, and lines along the inside and outside corners of large opaque regions. The patterns are sensitive to effects such as dose, focus error, lens and illumination aberrations, and flare. The patterns were projection printed at λ = 436 nm, NA 0.28 and σ = 0.7 on doped and annealed polysilycon, thin polysilicon and titanium coated wafers. The patterns after etching were automatically probed, analyzed and plotted on a flexible probing system. Significant flare and moderate linewidth variations across a die were observed. The exploratory focus targets gave greater sensitivity than the standard linewidth structures but were also strongly affected by exposure and bias both in mask making and in pattern transfer. A test pattern with programmed defects and an associated model for quantitative interpretation gave promising results for electrically monitoring the printability of defects.

Photolithography has been used for manufacturing LSIs for a long time and it also becomes a key technology for submicrometer VLSIs. Three dimensional photolithography simulator, RESPROT, has been developed, and reported its concept and application data at the last SPIE's symposium, Vol : 922-02. In this paper RESPROT was improved to simulate optical aberrations, proximity effects and repeatable reticle defects which are remarkably important process factors on submicrometer pattern transfer. And resist pattern printability or fidelity were studied. At first three dimensional simulator, RESPROT, was improved for the quantitative calculation of higher order reduction lens aberrations and contrast enhance layer, CEL, for resist images. It was confirmed by the calculations that there is existent sixth order aberration and critical dimension loss was decreased by using of CEL. For proximity effect it is slightly improved by higher numerical aperture, NA, but resist images are deformed on defocusing. Printability of submicron reticle defects are depend on the defect type ; clear or dark defects. Both defects decrease imaged resist sizes on defocusing, but these are transferred more clearly on higher NA imaging. Also reticle defects are printed with more faithful by optical aberrations.

The effects of object-to-image vibration are studied with the purpose to quantitatively define the maximum acceptable vibration level as a function of resolution, linewidth control, exposure and focus tolerances, proximity condition, and illumination coherence. The exposure-defocus trees and windows of optically projected images are used to derive the vibration tolerance as well as the optimum exposure as a function of vibration amplitude. Dimensionless variables for resolution, focus and exposure are used so that the results are universally applicable for any given diffraction-limited lens system.

We address here the optimization of the photolithographic process for submicron design rules. We describe a new method to determine the exposure-focus window corresponding to a user-specified design rule. The change in size and position of the window with changes in design rule, and in process and equipment variables, is the basis for optimization. We describe the implementation of our optimization approach in the MONO-LITH 11/88 Workstation; a computer aided engineering tool that standardizes data acquisition, analysis and presentation, independent of metrology or sampling approach.

The new evaluation method has been developed for the practical resolution of reduction optics by applying the quantitative value of resolution index and the accurate evaluation results are obtained. The quantitative value of Resolution Index is defined as the controllability of critical dimension and this value for the arbitrary pattern represents the resolution of reduction optics. The critical dimension of the resist pattern is altered by the focus position and exposure dose and the characteristic curve is obtained by this relationship. This characteristic curve will also express the allowable range to the defocusing and exposure dose for the critical dimension control and this range is determined by the resolution capability of given optics. The concept of this method is that the exposure and focus tolerance are simultaneously determined by defining allowable CD tolerance of the arbitrary pattern and the Resolution Index is obtained by calculating the defined area on the two dimensional coordinate of focus-exposure dose plane. The calculation of Resolution Index is very simple because the relationship of focus-exposure dose is expressed by the simplified equation. In this paper, the basic concept of this quantitative evaluation method; characterization of CD curves; application to the full exposure field; detail of the calculation procedure for the resolution index; practical resolution limit will be discussed.

The design of a manufacturable, cost effective one micron lithography process requires locating and optimizing the largest manufacturing window of focus, exposure, and bias for each critical cd layer. Special attention must be paid to the resist and thin film coupling effects on critical dimensions and resist clear point, Eo. Similar characterization and optimization must be performed for registration control as well. This paper explores optimization techniques that proved successful in determining the process parameters which yield the best manufacturing tolerances. Distribution centering techniques will also be discussed as well as the key approaches to implementing a process which does not require any lot dependent adjustments or test wafers. The case for non-adjustment processing can not be overstated for if the process is constantly "tweaked" for exposure or offsets, stability and control is impossible and so is low cost, high output production. The steps to use in designing, characterizing and controlling a no-adjust lithography process are outlined in this paper.

A high contrast single layer i-line resist process was developed using Aspect Systems 9 resist for a GCA 5X wafer stepper equipped with a Tropel 20 mm diameter, 0.40 NA lens. The process development sequence is described, where surface response experiments are used to evaluate the effects on resist contrast of process factors. These factors include prebake time and temperature, post exposure bake time and temperature, metal-ion free developer concentration, develop time, and develop temperature. Contrast is found to depend upon standing wave interference. Results of a factorial comparison of the single layer process for critical dimension control with and without CEM 388 made for site-to-site and wafer-to-wafer variations yielded similar results. SEM micrographs for the optimized single layer process illustrate near verticle resist profiles with good depth of focus latitude, and agree well with PROSIMulations except under large defocus conditions. The optimized single layer resist process has been employed to evaluate a new Tropel 20 mm diameter, 0.40 numerical aperture, i-line lens with good results.

A new optical stepper, NSR-1505G6E, has been developed and it will be one of the most promising exposure apparatuses for half-micron lithography in the next era. g-line is used as an exposure wavelength, and the 5X-type projection lens has 15 mm by 15 mm field size. Its N.A. (numerical aperture) is 0.54, the maximum value among the 5X-type g-line projection lenses ever announced. First, the capability of high N.A. is demonstrated by a theoretical approach. Conventional optical theory says that resolution power is proportional to λ/N.A., but multi-reflection within the photoresist is not taken into consideration. To investigate resolution power based upon resist patterns, one of the most powerful approaches is "the vector model" developed by M. Yeungl in 1988. A similar simulator has been developed and has been used to confirm that a high N.A. lens has high resolution power even with the multi-reflection effect within photoresist. Secondly, performance of the projection lens is shown on experimental data basis. Advantages and disadvantages of the projection lens are discussed in comparison with an i-line lens with same resolution power. Investigation on N.A. is also made, comparing resist patterns exposed by some lenses with various N.A. values. Finally, the total system, including such new functions as chip-leveling and auto-focus, is described.

In order to meet the production requirement for higher integration, g-line lenses of higher NA have been developed. It is generally believed that to enlarge both NA and image field is difficult. The maximum image field achieved with a high NA lens has so far been 15 mm square and the maximum NA of a wide image field lens has been 0.35. Canon has recently developed a new g-line lens of higher NA and larger image field. Results of evaluations so far show that the newly developed lens is the best-suited of any yet made for mass-production of 4M DRAMs. This is the 5x reduction lens with a numerical aperture of 0.45 and a field size of 20 mm square (28.2 mm dia.). In this paper, the performance of this lens is discussed, and SEM resist profiles produced by the new lens are shown. The last section gives an overview of progress in development of projection lenses in terms of the amount of information available, and also discusses the possibility of a lens for 16M DRAMs being developed.

As the semiconductor industry presses toward smaller and smaller feature sizes, new lenses with higher numerical apertures are being introduced. As more chipmakers purchase i-line steppers, the resist makers have developed new resists to exploit them. This paper will show experimental results from two new lenses offered by GCA using new, as well as established, i-line resists.

I-line lithography offers the capability to achieve half-micron integrated circuit design rules. Such design rules require very good optical performance matched to resist process technology. Overlay performance at these design rules is also critical wit capability needed in the 100-150 nm region. Resolution and usable depth of focus (UDoF) need to be evaluated concurrently and are influenced by the lens design (wavelength, NA, focal plane deviation) as well as the resist processing technology. A simple model is presented to describe these inter-relationships. Experimental comparison of UDoF performance from 0.7 μm to 0.5 μm resolution is presented for several resist processes showing UDoF performance of greater than one micron for half-micron features. Half-micron capability on IC film stacks and topography with good UDoF is also demonstrated. The PAS 2500 alignment and positioning systems are compatible with the overlay requirements for half-micron design rules. Improved performance of the key subsystems is reported; notably the reproducibility of the phase grating alignment system and the positioning accuracy of the three axis XY stage system. The resulting improved single machine overlay from 37 steppers is 120 nm 3 σ. Improved machine to machine matching will be reported based upon the reduced single machine overlay, smaller lens distortions and an improved matching procedure. The distortion data of 19 Zeiss 58 lenses will be reported in terms of the difference between all possible pairs of lenses. Multiple machine overlay data will be given and the improvements from lens selection demonstrated.

The need for large substrate, high-throughput, lithographic tools has led to the development of the MRS 4500 PanelPrinter. This new stepper combines image field stitching technology with a long-travel, air-bearing stage to achieve four micrometer (μm) design rules over a 450 by 450 millimeter substrate. In order to produce sufficient throughput, the PanelPrinter utilizes a novel architecture which employs multiple optical projection systems to create a single, large-area image. This paper presents a description of the PanelPrinter's system architecture and provides performance data to illustrate resolution, image distortion, and overlay performance on large-area images. Special emphasis is given to the on-board metrology systems which provide the stability and precision required to perform large scale lithography.

Achieving a proper focus setting in sub-micron optical lithography is extremely important. A technique has been developed for the determination of optimal focus setting on production lots. Using aerial image simulation the dimensions of the test structure studied have been optimized. With a simple optical microscope a judgement can be made about the focus setting. As the depth of focus becomes comparable to the resist thickness, it was observed that different methods of focus determination do not give the same results.

A series of alignment experiments was carried out in order to specify and optimize the design parameters for the site-by-site alignment marks used in an advanced single polysilicon bipolar process (ASPECT**). An Ultratech model 900 projection stepper which employs a dark field alignment system was used as the exposure tool. Alignment evaluations were performed for all masking levels of the ASPECT process with special emphasis placed on polysilicon and metal masking levels. The use of optimized targets resulted in a drastic improvement in registration accuracy: alignment statistics collected on a large number of samples with optimized targets for polysilicon to isolation alignment indicated a registration accuracy of +/- 0.27 um (2 sigma), whereas standard targets yielded +/- 0.50 um (2 sigma). For alignments of metal levels the values were +/- 0.40 um and +/- 1.0 um, respectively; thus the use of optimized targets allowed on reflective layers an alignment accuracy equal to or better than the stepper capabilities published by the manufacturer: Ultratech specifies an overlay budget of +/- 0.40 um (2 sigma) for a single stepper. This improvement in overlay accuracy translated into a large decrease in number of reworks and a significant improvement in product yield.

Laser alignment systems are widely used in optical lithography for LSI production. They, especially the dark field type, have high signal-to-noise sensitivity to step height or grating line alignment marks, because of a sharp and high intensity spot owing to coherency of laser beam. The coherency often distorts alignment signals according to a little change of interference condition caused by variation of alignment marks. This phenomenon causes overlay accuracy to Al layers to deteriorate. This is because that the Al layers sometimes have a cracked or granulated rough surface and asymmetrical alignment marks which produce noise and deformed signals. In order to analyze and optimize these effects, a computer program has been developed. Laser Step Alignment (LSA) is used as a model system. With this program, it is possible to calculate the alignment signal from 3-dimensional alignment mark and resist profiles. Then, the overlay errors are analyzed from the simulated signals at various detection levels. In one optimized case, when a concave mark is used, the lower detection level should be selected to minimize the influence of Al asymmetrical coverage. This case corresponds to many experimental results.

The natural progression of today's semiconductor industry is toward smaller geometric features and registration requirements. Typically, this progression results in high capital equipment investments, along with a large capacity reduction per investment dollar for most lithographic exposure processes. One major cause for the capacity loss is the industry's willingness to migrate from full-field scanning projection printers to a lower throughput field-by-field alignment step-and-repeat exposure system. Standard Microsystems Corporation (SMC) sought to achieve higher performance on its scanners without compromising throughput. The original goal at SMC was to improve Perkin-Elmer's Micralign 641 HT machine-to-machine registration specification of ± 0.30 micron to less than ± 0.25 micron. With this in mind, we set out to investigate the true alignment and registration limitations of a Micralign Model 600 HT Series Projection Aligner. Although SMC was apparently successful at matching two Micralign 641 HT systems to ± 0.25 micron by manually reading verniers, this technique proved to be time consuming and prone to human error. Electrical probing of wafers was considered, but the special masks and processing steps and its destructive nature were considered undesirable. For this study, an automatic optical overlay measurement system was used to optimize overlay on the SMC Micralign systems. The results were enlightening. The specified overlay of ± 0.30 micron for 98% of the data improved to better than ± 0.25 micron, 3 sigma. These results were achieved without the use of Automatic Magnification Compensation (AVM/AMC). We also discovered that many otherwise transparent mechanical/optical anomalies, such as contamination and scan interference, could be readily identified. Experimental data is presented and the beneficial application of this technique to a production process is discussed.

We are developing a projection moire alignment technique in order to obtain an alignment accuracy of better than 0.1μm(3σ) using simple alignment optics in a projection photolitographic system.[1] Chromatic aberration between the exposure light(g-line) and the alignment light(He-Ne Laser) was compensated using fundamental moire theory without compricated optical unit. The beam pointing stability of He-Ne laser has affected an alignment accuracy. By taking reference beam befor incident on the wafer, we have removed this problem. We obtained an alignment accuracy of 0.07μm in resist mark and 0.10μm in Si etched mark.

A dual measurement interferometer has been developed specifically for wafer stage metrology. The interferometer concurrently measures both linear and angular displacement. It is designed to conserve space, minimize stage mass, eliminate heat sources and provide high resolution and slew rate.

One of the major challenges facing the semiconductor industry today is to be able to respond to an increasingly competitive environment. One effective strategy is to reduce costs through advanced operation, maintenance, and information flow techniques for manufacturing equipment. This approach can yield very beneficial results for manufacturing economics.

Application Specific Integrated Circuit (ASIC) manufacturing, characterized by fast cycle time, small lot sizes and an abundance of products and process flows, creates challenges not found in a dedicated product line. The continuous addition of new devices into the wafer fab creates problems such as reticle storage, critical dimensions specifications and stepper set-up. One of the stepper set-up problems is stepper job file generation. Each device run on a stepper requires several job files to be created. The number of job files created for each device is dependent on the manufacturing environment, the alignment scheme, process flow and stepper used. There can be as many job files as there are layers for each device. Data from the reticle layout engineer is entered into the stepper by the process engineer to define the stepping array, location of alignment marks, framing blade settings and other exposure related information. This process is time consuming and removes a machine from production while job files are being written and tested. In an ASIC environment where several new devices start each week, creating stepper job files is a major problem impacting: productivity, cycle time, process yield and the rework rate. A solution to all of these problems: automate the stepper job file generation process. Our stepper job files are created automatically by computer when the reticles are tooled. Stepper job files are ready before the masks arrive and they are always correct. The result is improved productivity and process yields with reduced reworks and prototype cycle times. A description of this automated approach and the necessary items to implement this technique is presented.

With the developments in the techniques of artificial intelligence over the last few years, development of advisory, scheduling and similar class of problems has become very convenient using tools such as PROLOG. In this paper an expert system has been described which helps lithographers and process engineers in several ways. The methodology used is to model each work station according to its input, output and control parameters, combine these work stations in a logical sequence based on past experience and work out process schedule for a job. In addition, all the requirements vis-a-vis a particular job parameters are converted into decision rules. One example is the exposure time, develop time for a wafer with different feature sizes would be different. This expert system has been written in Turbo Prolog. By building up a large number of rules, one can tune the program to any facility and use it for as diverse applications as advisory help, trouble shooting etc. Leitner (1) has described an advisory expert system that is being used at National Semiconductor. This system is quite different from the one being reported in the present paper. The approach is quite different for one. There is stress on job flow and process for another.

Experimental results of a high-resolution holographic imaging system using wave-front conjugation are demon-strated. Using an Argon laser operating at 488 nm with a coherence length of one meter, a hologram is formed from the information on a master mask that is imaged onto a high-resolution recording material with a photographic quality multi-element lens. The hologram is illuminated with a spatially and temporally filtered beam of light centered at 488 nm from a Hg arc lamp to construct the image of the master mask. The aberrations introduced by the imaging lens are eliminated through the process of reverse ray tracing back-through the original lens. The use of a spectrally-broad construction source reduces the effects of optical noise generated in the optical train during the construction of the image. Diffraction limited images are experimentally observed throughout the illuminated field.

In a previous study [1] a new method was developed to describe the effects of defocus on an optical lithographic process. The interaction of the aerial image with the photoresist was described mathematically in order to determine the features of the image which are important in determining lithographic performance. The slope of the log-image was determined to be an appropriate metric of aerial image quality. By calculating this log-slope as a function of defocus, rigorous definitions of both depth-of-focus ( DOF) and resolution were given. The DOF, for a given feature size, can be defined as the amount of defocus for which the log-slope of the aerial image remains above some minimum value. The minimum value of the log-slope which gives acceptable process latitude is determined by the properties of the photoresist process. This paper discusses the important properties of a photoresist and how these properties affect DOF. The primary parameter lithography model PROLITH [2] is used to investigate how various process parameters change the response of the lithographic system to focus. The results are compared to the log-slope defocus curve to determine the minimum acceptable log-slope for the modeled system. Finally, experimental linewidth data was collected as a function of focus and exposure using electrical linewidth measurement techniques. This data is compared with both the modeled data and the log-slope analysis.

A resist evaluation was performed to select a high resolution positive resist which would meet all the requirements for 1µm CMOS technology. The leading resists in the 1µm evaluation were further evaluated as replacements for an existing production gate imaging process employing CEM. The methodology used in the evaluation will be described with particular emphasis on gate and metal imaging.

With MOSFET technology the dominant device used in VLSI circuits, as well as the continued reduction in device dimensions, a particularly challenging lithography goal will be the patterning of submicron CMOS gate electrodes on grainy, reflective polysilicon films. This paper will provide a detailed investigation of the microlithographic optimization of such devices. Utilizing a 1:1 0.40 numerical aperture (NA) broadband lens stepper system with both high contrast mid-UV and near-UV resists, we investigated extending lithography to near resolution limit capabilities on this system by patterning CMOS gate electrodes on grainy, reflective polysilicon films for line/space dimensions of 0.8, 0.7, and 0.6 μm. Electrical test results of patterned substrates are presented to indicate the practical, ultimate resolution and process latitude of this system. The electrical results include detailed electrical data which represent the entire lithographic process including dry etch. The data was taken at 50 sites over the entire optical lens extent. Selected scanning electron micrographs of resist profiles are included to support these findings. The effects of focus and exposure variations are presented to illustrate the feasibility of implementing this system for process manufacturing in CMOS technology.

This paper propose an user characterization technique for optical steppers. The purpose of such technique is to understand and to control those factors which affect stepper performances when fabrication of structures close or below the resolution specifications of the optical system is required. Two main factors are considered to limit stepper capabilities in submicron lithography: the imaging optics as a transmitter of high spatial frequencies with certain image intensity profile and the photoresist as a receiver of this optical information. The reported experiments use as the resist receiver a positive resist with a basic component added (0-4% by weight) in order to increase the resist transfer function and to decrease the K factor (from Rayleigh two points resolution criteria for isolated patterns, R = Kλ/N.A.). From spectral absorption of the in-house doped resist, we found that doping component concentration makes the process suitable for high resolution/ minimizing reflectivity/stepper target recognition. The imaging optics factors, a 0.31 N.A. /10X and 0.30-0.40 N.A./5X g-line steppers have been tested in the reversed resist process for linewidth change with defocus for various first/second exposure doses;exposure-defocus window;depth of focus for tangential/sagittal 1 micron lines/spaces patterns;lens capabilities to reproduce over all 10x10 mm field horizontal and vertical 1 micron lines. The latter is a useful indication in mixing-and-matching two steppers. Minimum 0.5 micron isolated patterns were reproduced in the reversed process, simultaneously in opposite polarities. The imaging optics was tested also for third order distorsion and errors of 50 ppm at optical field edges could be simply detected.

In a Junction Charge-Coupled Device (CCD) electrons can be trapped in parasitic energy wells near the edge of the channel. This results in a low transfer-efficiency. These parasitic energy wells can be diminished by a change of the confinement. This was materialized by simply modifying the shallow p-type diffusion mask. The result was a reduction in the transfer-inefficiency by a factor of ten.

A novel etch masking process for orientation dependent chemical etching of silicon is presented. This process is based as well on the conversion of a surface relief pattern into a surface concentration pattern using ion implantation technique as on the differential etch rate of the heavily doped p+ silicon. The ion implantation process allows direct transfer of the photoresist pattern to the substrate. The improved etch control enables wet chemical etching of submicrometer structures. The different behaviour of the heavily doped p+ silicon versus various etchants renders possible to obtain patterned image reversal.

In this paper, we report the results of ANKAN, a computer program written for the exposure and development of positive photoresist using matrix formulation. Exposure process is modelled using diffraction limited optics and partially coherent light source with a wavelength λ. The program uses string model for resist development. The results of theoretical computations are reported and are found to be in good agreement with those of previous workers. In addition, the program can also be used for computing edge threshold for linewidth measurement and predicting the shape of aerial image for two dimensional objects.

An analytical model for 2-D non periodic aerial image simulation is proposed for rectangular and circular mask structures. As applications we demonstrate that better resolution in contact opening is acheivable if near circular pattern are designed at mask level. Examples of optimized aerial image for ODB tapering method are shown. Influence of defects on the reticle are considered.

The "wave-guide" model which can be used to study the light scattering problem in optical alignment, has been extended to include illumination with arbitrary polarization and large numerical aperture (NA) lens. An existing simulator has been enhanced to calculate the images of different structures observed under different alignment schemes. It is shown that variations in the polarization conditions and topography may drastically alter the alignment results.

A commercial Lambda 248 L excimer system has been equipped with a newly developed wavemeter. The self-controlling laser system can be stabilized at a pre-selected wavelength in the range of 248.180 to 248.580 nm with an absolute accuracy of ± 0.001 nm. Due to the deep-UV stepper optics the wavelength can be adjusted absolutely to comply with different optics. The modified 248 L-system can be operated at an average power of up to 5 W. Experimental results are presented.

We present the design concepts and performance data for a spectrally-narrowed, wavelength-stabilized KrF laser that meets the specialized requirements of a deep-UV light source for a stepper. Features include: (1) compact size (54cm W x 45cm H x 128cm L), (2) 3 Watt average power with spectral width of 0.003 nm FWHM, (3) long fuel gas life from premixed bottles attained without the need for a gas reprocessor or boost gas, (4) fast and precise wavelength control system that maintains a user-specifiable wavelength to within 0.0005 nm anywhere in the KrF tuning range, (5) microprocessor control, including SECS or custom control protocols, (6) robust, modular internal design for rapid repair, and (7) engineering and packaging that meets semiconductor production tool safety requirements.

In this paper, the performance characteristics of the INDEX 300 lithographic industrial excimer laser are presented. The laser operates with a linewidth of 3.0 pm and spectral purity of 94%. The wavelength stability is better than ± 1.0 pm over short and long term operation in both continuous and intermittent modes. By means of a novel technique, an absolute wavelength standard is used as a reference for wavelength stabilization. The advantages of this approach to wavelength stabilization are discussed.

The Micrascan I system, a new 0.5-micron lithography tool, employs a 'step-and-scans concept that combines the virtues of ring-field scanning systems with those of conventional step-and-repeat systems. Individual exposure fields on the wafer are sequentially scanned past the arcuate image field of a catadioptric projection optics system with a 4:1 reduction ratio. The reticle is scanned in precisely coordinated fashion by a separate stage. Exposure illumination is provided by a long-life mercury-xenon arc lamp. The concept enables very large field sizes and high system productivity.

An excimer laser stepper having a flat image field was developed. Using a frequency stabilized, narrow band KrF 248nm laser (FWHM≈3μm), this stepper can print uniform patterns under 0.5 μm over a 15 mm square field. When the excimer laser output power is 2W, illumination intensity is more than 40 mW/cm2. Illumination uniformity is within ±2.5% over the field and exposure energy density is controlled within the precision of ±1.596, using an integrator sensor and a pulse number controller. A chromatic projection lens is adjusted so finely that aberrations are removed. In order to get imaging stability, a lens controller is used. An alignment system is off-axis, using He-Ne laser light spots and diffraction light detectors. By employing the EGA (Enhanced Global Alignment) method, overlay accuracy better than 0.18μm was obtained. Because the alignment detection light is insensitive to the resist, any types of resist may be used. The developed system has all the basic functions of a stepper. It can be used to improve DUV process and to fabricate sub-half micron devices in a laboratory.

We have constructed and tested a lens for use with deep ultraviolet photolithography systems. This broadband lens may be used with a mercury arc lamp as the illumination source. It is found that this system is an economically sound approach to the implementation of sub-half micron photolithography

The relevant characteristics of a 248 nm line-narrowed excimer laser have been measured. The laser's spectral width is 1.6 pm FWHM and its coherence length is 38 mm. The effects of this line narrowing on the optical components in illuminator systems have been studied. The results from several light pipe systems and a diffuser plate system are presented. The effects of external cavity collimation on the GCA-TROPEL BOLD illuminator are also presented.

PR1024MB photo resist, one of the novolac based positive type resist for i-line lithography, showed nonreciprocal behavior under the condition of high power KrF excimer laser exposure, and a negative type reaction was induced. This reaction was accelerated by an additional reversal bake and a flood exposure, which were well known technique as image reversal process for g-line and i-line lithography. In the image reversal process with KrF excimer, it was observed that the sensitivity became higher as the reversal bake temperature increased. Furthermore, the change in resist pattern profile from undercutting shape to rectangular form was observed with the increase of reversal bake temperature. However this image reversal process required a higher exposure dose because of its nonlinear reaction, the resist pattern profiles obtained were much better than the triangular shape formed by the normal positive type reaction. The dependence of resist pattern profile and sensitivity on flood exposure wavelength and energy was quite small. It was concluded that the key parameter of this process was the reversal bake temperature, and fine patterns less than 0.5 um with steep sidewall were obtained by optimizing the process conditions.

A 'Microstepper' is being constructed at U.C. Berkeley as a flexible lithography tool to study the possibilities and problems of photolithography using refractive optics with a KrF Excimer laser light source. This paper will present data on the performance of the optical system. 0.3μm line/space elbows were clearly resolved in thin MP2400 resist. Single-pulse, speckle-free images were achieved by modifying the condenser design and using a high-sensitivity chemical-amplification resist. Optical noise was found to be a problem. We were able to substantially reduce this noise by using a moving diffuser. Alternatively, incorporating a light pipe into the condenser design was also found to reduce noise in the image. Data are also presented on the use of air probes as height sensors for this demanding application (the Rayleigh depth of focus for a 0.6 N.A. lens at 248nm is 0.7μm). Data showing the sensitivity, stability and reproducibility of some different probe designs is shown. Finally, we present the design of the microstepper which is currently in the final stages of assembly.

The life of optical lithography has been prolonged by introducing Deep UV sources and its limit seems to be in the sub-half-micron to quarter-micron region. In Deep UV lithography, KrF excimer laser lithography is the most promising technique for the next generation. We have evaluated the resolution capability of a KrF excimer laser stepper using a conventional novolac-type resist. With this combination, it has been shown that an alkaline treatment is an effective method of improving the resist profile, which means that a conventional resist is applicable to excimer laser lithography. The mechanism of resist profile improvement is also discussed. We assert that a kind of chemical reaction is caused during the PEB step under certain conditions.

A new method to fabricate the gratings useful for integrated optical circuits (IOC) is described. The method combines optical projection lithography with spatial filtering. The projection lens uses nearly coherent illumination. Only the two first orders diffracted by the mask grating are allowed to pass through the lens. This produces a grating pattern in the image plane that has a high contrast (near 100%), a large depth of focus (approx. 13 μm), and a period half of what would have been obtained in normal imaging. Thus, with a 5X reduction lens a 5 μm period grating on the mask produces a 0.5 μm period image on the wafer. 0.5 μm period gratings are useful for fabrication of Bragg filters in dielectric film waveguides. Gratings of different periods, sizes, locations, orientations, and configurations (chirped, phase shifted, etc.) can all be produced on the same chip with a single exposure. Using a deep UV lens with a line-narrowed 248 nm KrF excimer laser for illumination, we printed 0.5 μm period gratings in an oxide layer on 10 mm X 10 mm silicon chips. Because of the high contrast, the photoresist patterns had very high quality. In addition, the extended depth of focus was observed. This method is primarily useful for patterning periodic structures. Lenses designed for this system could be made with large numerical apertures and/or image fields, and be able to pattern gratings of 0.23 μm period for semiconductor waveguide devices. This technique opens up the possibility of high-volume production of IOC chips with Bragg filters using standard IC fabrication facilities.

The lithographic processes involved in the fabrication of thin film heads are discussed. Because of the intrinsic features of huge topography and high aspect ratio, process requirements for film heads as opposed to semiconductor devices are identified. A process of conformal thick resist coating across a large step height is developed by optimizing resist coating parameters and the use of a resist with higher viscosity. Also planarization techniques are discussed with a specific example illustrated to planarize the topography associated with the fine and large cross-sectional coil pattern. Alignment accuracy proves to be the most critical parameter for further advances in the film head technology and several strategies are proposed to improve the alignment capability.

Conventional UV photolithography has been applied to the replication of glass compact disk structures for low cost fabrication of grooved glass substrates and ruggedized CD-ROM and CD-AUDIO disks. Unconventional pattern generation and replication strategies have been employed to produce high quality compact disk structures in soda-lime glass substrates.

A concept of fabricating high quality optical passive components directly into mask blanks is introduced to simplify fabrication processes. The concept is demonstrated by fabricating ion-exchanged wave-guides, diffractive binary amplitude and phase coded holographic optical elements, Dammann gratings and diffraction-free beam elements.

New optical lithographic technologies for 16 Megabit DRAM mass production will be required to generate workable 0.5 to 0.6 micron geometries for several of the device's levels. This is because current stepper and resist technology, while capable of resolving in this range of geometries, can not adequately do mass production processing at these levels due to optical limitations. These limitations have been described via the Rayleigh equation: W = k1 λ/NA