Competition is defined as a spirited, sometimes ruthless, engagement of rivals such as in a race, a match, or an effort by one person to sell goods or services to customers in the marketplace of another. Sound familiar? If you will bear with me for a few minutes, I would like to examine competitiveness on a more global basis with emphasis on the rules of the game. You may be thinking that more often than not the competitive arena is relatively small and far from global, and its consequences are singularly influential on a trivial document called the P & L. However, with the newly established freedom of a major segment of the world population and with the industrial capability formerly known as Communist moving into what has heretofore been "our" limited arena, the competition could get very brisk. Brisk, and perhaps ruthless, unless we work together to try to establish an international industrial policy that is truly based on equality of competitive opportunity for all.

The phase shifting mask is a two-layer engineered structure that improves the resolution of microlithography by between 25% and 100% by using interference to cancel some diffraction effects^1,2. Although invented in 1980, this technology has only recently become prominent, mostly because of its potential to produce 16 Mb and 64 Mb DRAM with current i-line projection tools^3,4. In the long run, phase-shifting masks may be used with deep-UV steppers to make structures with critical dimensions down to 0.15 micrometer.

In this Era of shrinking design constraints, two (2) of the many concerns facing E-Beam mask makers are: (1) data snapping on design grids smaller than E-Beams can fabricate and (2) dense designs with submicron features forcing smaller beam steps which radically effect write time and throughput. The Philips/Cambridge Vector Scan E-Beam provides novel approaches to meet Western Digitals 0.05 (mu) gird resolution, yet still maintain reasonable throughput. Using CATS Transcription, designs are fractured into two (2) or three (3) distinct patterns called a Bulk/Sleeve or Bulk/Double sleeve technique. The Bulk pattern comprised the majority of the pattern and is written at a large beam size. The Sleeve is a border pattern around all geometries written at a much smaller beam size. This combination allows throughput because of the large beam sized Bulk pattern yet gives high resolution, and edge acuity with the small beam sized sleeve pattern. By combining a Bulk and Sleeve pattern exposure matrix with a smaller bias reduction a sleeve of 1.2 (mu) is currently utilized in production. Additionally sleeve size reduction reduces the complexity of the C Format trapezia and virtually eliminates data snapping. An interesting by-product of this Bulk/Sleeve technique allows multiple exposures for the different patterns to help alleviate proximity exposure effects in extremely dense designs.

Mask ROMs are one of the most advanced devices commercially available today. 4 Mbit DRAMs are just coming to the market, where 32 Mbit and 64 Mbit Mask ROM products also are entering the market. The code mask of Mask ROM consists, of a random distribution of repeated cells on a constant grid, so it is difficult to use conventional e-beam data compaction techniques for the code mask level of a Mask ROM. As mask ROM capacity and their mask exposure data volumes increase, the e-beam data processing time for these devices also has increased even with enhanced computational speeds of new mainframe computers. To overcome this problem, we have adopted a new e-beam data compaction technique. With this ne technique of a data compaction, data conversion times on a mainframe computer (ACOS) are substantially shorted and data volume is reduced by as much as a factor of one hundred. Using this new data format, the data volume of a variable shaped vector scan e-beam exposure system became far smaller compared with a spot beam raster scan e-beam system.

The development of a high resolution laser lithography tool which utilizes a 0.6 NA 20X reduction lens, 364 nanometer exposure source and a four pass writing strategy provides the opportunity to evaluate an optical lithography tool for mask making that has removed some of the inherent complexity associated with E-Beam exposure systems. No high vacuum system is required and there is no need for grounding of the substrate to dissipate the charge induced by the exposure system. The plates are not carried in individual cassettes and therefore the system performance is not subjected to additive tolerances associated with this method of handling. Freedom from the adverse affects of these sub- systems should appear in the form of a product performance in the areas of registration and defect control. The production capability of the system is analyzed to determine what type of routine performance can be expected in terms of resolution, linewidth control, registration, defect additions and average print times. The primary goal is to determine if the performance in these areas is capable of meeting reticle requirements for 16 megabit design rules which are 0.5 micron geometries on the wafer. The evaluation revealed the basic capability exists, however, additional work must be done in the area of resist/process to achieve viable production performance for linewidth control.

Proximity effect correction is necessary to fabricate masks with 0.25 micron design rules using electron beam lithography. The GHOST technique of proximity correction has the advantage of no pattern preprocessing and is easily implemented on a raster scan system such as MEBES. Recent results show proximity corrected features at 0.3 micron. To minimize constraints on the resist characteristics, such as the Srg ratio, global sizing of patterns has been investigated and found to provide an additional degree of freedom to control sensitivities and process latitude. Simulation and experimental results will be presented to demonstrate the use of GHOST and sizing for 1X mask making, including discussion of some of the relevant issues and tradeoffs.

An alternative exposure strategy has been developed for the MEBES electron beam exposure system that enhances throughput and improves registration accuracy of 5X reticles. The Multi-Phase Virtual Address (MPVA) exposure strategy utilizes a voting technique to minimize scan related distortions while increasing writing address size thus decreasing the write time significantly over present methods. The resist chosen to evaluate this writing strategy was Hoechst AZ5206. The technique was evaluated to determine performance capabilities for mask registration, linewidth tolerance and uniformity, scan related error reduction and throughput improvements.

As competitive pressures continue to test the viability of every manufacturing enterprise, an ongoing program of operational improvement is essential. Continuous Flow Manufacturing (CFM) is a methodology created and practiced by IBM to meet this need. CFM combines the elements of total quality control, total people involvement and the elimination of waste to insure continuous attention to enhancements of manufacturing efficiency. This paper provides an overview of CFM and suggests six generic areas of every manufacturing line where the CFM approach can be used. The CFM methodology has been applied to an IBM internal business unit that manufactures photomasks used for semiconductor production. In 1984, serviceability and quality measurements in ths business unit were unacceptably low; and business measurements were nonexistent. CFM provided the framework for dramatic operational improvements in this business unit. Today, serviceability in the 90% to 100% range is routinely achieved. Delivery times have been more that cut in half, while superlative quality mesruements have been attained. Finally, cost reductions have been realized in an environment of ever-increasing technological challenge. Plans for future improvements using the CFM method are in place. The goal of all manufacturing endeavors has always been, and still is, ongoing operation improvement. CFM offers a structured methodology for pursuing this goal.

Most photolithographic masks are made today using technology that has charged little from that licensed to the industry by Western Electric in 1975; even X-ray lithographic masks have changed little since then. This at first sight is extraordinary given that this same period spans the evolution of DRAMs by three orders of magnitude (from 4Kbit to 4M-bit) and the reduction of feature sizes nearly one order of magnitude.

The challenges of producing 16M, 64M and perhaps 256M DRAMS in the 90's, requires mask/reticles with better specs than that can be provided by present generation systems. Overlay error, pattern size uniformity (CD control), pattern placement error (stitching) and patterns linewidth (MFS) have to improve by at least a factor of two over existing systems. In addition, the time required to write these patterns has also become intolerable on present systems. A faster, higher resolution and more accurate system is needed and EBES4 meets all these requirements. From minimum feature size standpoint, EBES4 provides 1/8 (mu) writing spot and can write (MFS) from 1/8 (mu) to 1/2 (mu) , depending on resist sensitivity, with high accuracy. The three tiered deflection system optimizes both speed and accuracy, so that 500 MHz data rate can be achieved while not introducing writing error larger than 0.05 (mu) (3(sigma) ). EBES4's proprietary TFE cathode provides not only high current (250 na) so that 500 MHz rate can be achieved with 3 (mu) C/cm² resist but also very long lifetime of 8,000 hours. The vector deflecting and raster fill-in strategy minimizes non-writing overhead time as in conventional put raster system. Attention has been paid to kinematically hold the mask and use temperature insensitive material for the metrology sub-system of EBES4. Overlay accuracy can be improved by a factor of two over MEBES3. Topics covered are DRAM Reticle Requirements, EBES4 system error budget, EBES4 Electron optics and beam stability. Stage and Meteorology design considerations are outlined. Writing strategy and Software requirements are discussed. Finally we discuss hidden pattern writing issues such as any angle pattern, corner radius etc.

Micronic Laser Systems AB, a developer of laser pattern generator systems based in Sweden, and Align-Rite Limited, a maskmaker based in the United Kingdom, decided in early of 1989 to cooperate on a project intended to utilize Micronic's laser scanner systems and Align-Rite's maskmaking experience to develop a system tailored specifically for 5X reticle manufacture. However, it was also found that the initial design was reliant upon a high degree of engineering support and was not well suited to a production environment. Based on the experience of Align- Rite; modifications were introduced into a prototype system that was installed at Align-Rite, Bridgend for evaluation. Continuing work by Align-Rite and Micronic identified areas where further design and process modifications were required in order to qualify the system as a heavy duty production tool. These changes are incorporated int eh production version of the machine, Micronic LSR-18, that is now being built in Sweden. The presentation intends to give the audience an understanding of the principal features and operational characteristics of the laser pattern generator, to discuss the results achieved thus far, and the process that is being used to generate 5x reticles. Finally, the future path and expectation of the project will be given.

This paper introduces a novel concept, 'chromeless phase-shifting', that eliminates the need for the use of chrome to form patterns in optical lithography. Chromeless phase-shifting uses 180 degree(s) phase-shifters on transparent glass to define patterns. The method relies ont eh destructive interference between phase-shifters and clear areas at the edges of the phase-shifters to define dark or opaque areas on the mask. Gratings sufficiently small will produce sufficient interference to completely inhibit the transmission of light (these gratings are thus named dark-field gratings). The combination of these effects makes is possible to form a wide range of patterns, from line-space patterns to isolated bright or dark areas. In this study, the lithography simulators SPLAT and SAMPLE were used to understand the principles behind this new scheme, and to verify various pattern designs. Simulation and experimental results are presented to demonstrate the concept.

Optical lithography has benefitted from progresses in wavelength reduction, lenses, resist systems, alignment, focussing, table accuracy, and insute metrology. As a result, the minimum feature size of integrated circuits has been reduced through many generations from 2 micrometers to 0.5 micrometers in manufacturing. There are many opportunities in improving the mask to help continue the progress. The image contrast can be restored by reducing stray light with mask anti-reflection coating at the absorber-air interface. Pre-distorting the mask against the distortion of the imaging lens can drastically improve the overlay performance. Adjusting the gray level or the feature size individually according to the pattern proximity environment can create a larger common exposure-defocus window for different feature shapes. Introducing phase shifts in the mask can simultaneously improve resolution and depth of focus, with the potential of a two-generation improvement with any given projection imaging equipment, provided the overlay capability is upgraded accordingly. In addition to describing and comparing these opportunities, the phase-shifting technology is given a special focus on the working principle, the different approaches to phase shifting, their imaging characteristics in terms of exposure-defocus diagrams, a systems view, and the scope of its development.

An algebraic image perturbation model is introduced which used the point spread function of the lens and the mutual coherence of the illumination to give insight into the interactions of auxiliary patterns with features for the phase shifting mask technology. The model is based on adding electric field contributions and the cross term is shown to characterize the dominant interaction as a function of the number of auxiliary features, the relative coherence of the illumination, and the spreading of the image of the auxiliary pattern toward the feature. Data on the point/line spread functions and the mutual coherence are given and used to verify the accuracy of quantitative predictions of the change in peak intensity for lines and contacts when nonprinting phase shifting auxiliary patterns are added.

A critical product property in the photomask industry is registration. Registration is a measure of how close features placed on the mask actually are relative to where it is desired for them to be located. Of course photomask customers, the integrated circuit (IC) manufacturers, want all features placed exactly where indicated by the circuit design. Consequently, the recent history in the photomask industry has been for specifications on registration to increasingly tighten. Today +/- .25 micrometers specifications are commonplace, and many customers are pushing for specifications of half this tolerance as they move into the production of chips of steadily increasing sophistication (eg. 4 and 16 meg devices). As registration specifications increasingly narrow, a serious question facing the photomask industry and their customers is not only whether the presently available lithography tools can conform to the tighter specs, but whether such specs are pushing or beyond present ability to measure registration. The present industry standard for assessing registration is the Ninon 2i; however, there ate two systems which are presently commercially available which claim to be superior to the 2i. One is Ninon's own successor system the Ninon 3i, and the other is the IMS-2000 made by Leitz. This paper discusses the findings of a study conducted by Du Pont Photomask to assess the relative accuracy and repeatability of each of these systems, and evaluates this information against the present trends in industry specifications.

Some problems encountered when measuring submicron CD patterns with white light conventional optical microscopy are presented. These are intended to support the development of a flexible software controlled new high performance optical system, allowing variation in the illumination degree of coherence, numerical aperture, threshold edge detection criteria, focus plane, giving the user opportunities to optimize the measurements. The paper will present results in measuring 0.90 to 0.65 patterns with opposite polarities in different measuring conditions from which an optimum 'operating point' can be defined considering either +/- 0.030 micrometers accuracy criteria or a +/- 0.020 micrometers reproducibility criteria. From such operating points an optimum numerical aperture and illumination aperture can be selected for a given edge detection criteria. NA selection for best resolution and accuracy is also function of the measuring plane position. For small amount of changes in focus plane, increasing the objective NA may actually decrease the quality of the image intensity profile used for measurement, resulting in decreased reproducibility. Such effects, reported also for submicron lithography imaging systems were observed. An increase in overall accuracy when focus plane is moved from the resist upper surface to the bottom resist surface (wafer surface) is observed for windows measurements.

A new reduced area electrically probed registration measurement pattern is described. The pattern is compatible with the Prometrix data acquisition and analysis system, and offers advantages over standard patterns in terms of patten area and versatility of use. The results of the application of the pattern to the measurement of reticle and stepper overlay are presented. With careful analysis of the data, inter- and intrafield reticle overlay errors are determined. Horizontal and vertical measurements of pattern placement within a single field and between fields showed an accuracy of greater than 67 nm and a repeatability of better than 14 nm (3 sigma).

Although full implementation of x-ray lithography as a production technology remains a few years in the future, there are now many world wide efforts to accelerate this introduction. Unlike other more common lithographic techniques, such as image projection, x-rya lithography requires the fabrication of a mask with a thick absorber to efficiently block the X-rays. This important distinction from the reticles used in wafer steppers requires a completely new approach to many of the techniques of mask making, including inspection and repairs. Focused ion beam systems have been suggested as a possible repair strategy, and a number of groups have utilized the inherent advantages of FIB methods to repair X-ray masks in the laboratory. Although FIB systems have achieve substantial acceptance in the photomask making community for repair of chrome masks and reticles, a simple reapplication of these systems to repair of X-ray masks will not produce the quality levels required in X- ray lithography. The purpose of this paper will be to review the primary technical problems in the repair of X-ray masks and to discuss the implications of these requirements on the design of an FIB system. The current state-of-the-art in X-ray mask repair will be reviewed and some unique results will be presented.

As mask and reticle designs continue to evolve in complexity and resolution requirements, maskmakers are investigating what advantages negative acting electron beam resists may have in meeting these requirements. One candidate is Poly (glycidyl methacrylate-co-3- chlorostyrene), GMC, which is an advanced negative resist used for the purpose of photomask fabrication. In this paper, a statistically designed experiment will be described in which GMC resist was evaluated for use on the MEBES system. Variables explored included exposure dosage, chrome etch time, resist descum and strip time. The effects of these variables on defect density, critical dimension (CD) size and uniformity will be presented.

The device density of integrated circuits has progressed with a 3 year cycle. The 256K DRAM was introduced in 1983 and the 4M DRAM will be in full production in 1990. The 5x stepper, which is the main lithography tool has progressed from G-line to I-line and it is said that the I-line will be able to achieve lithography for 16M DRAM's and some part of the 64 Megabit. We show the relation of this device trend compared with the required detection size of a particle and our particle detection system in TABLE 2.

Four years ago, we started pelliclization on both sides of masks using a pilot plant and shipped them for our customers for the first time. At that time, there were few pellicles that were usable, and we were only inspecting for particles. But recently, we understand the pellicle market in Japan to be approximately 20,000 pieces per month. It's convenient to the customer for them not to have to worry about the cleaning their masks during use because of particles being added to the plates. Therefore we believe the need for pelliclized mask will increase rapidly in the near future. On the other hand, specifications for 4M or 16M DRAM masks are very critical as shown in Table-1 and our customers are requiring the same particle size specifications under the pellicle film as the chrome defects on the surface of the mask. It's much easier to order such a mask than to make one evaluating a wide variety of pellicles and have developed adequate methods of mask cleaning and inspection.

E-beam lithography has for some years been the leading edge lithography tool in many advanced technology programs and in research and development. More recently it has found its place in the production of mask and reticles for a wide variety of products, both at 1X and for magnification applications. The traditional systems have utilized spot beam optics with relatively low current density and, whether raster scan or vector scan systems, have resulted in many cases in long job times for some of the most advanced device requirements. The inherent cost benefits of direct write lithography have seldom been met with such tools due to the 'cost per level' factor resulting from a low throughput approach. During the last few years variable shaped beam systems have become commercially available and these systems have substantially improved system throughputs for most applications and have, for the first time, enabled E-beam lithography tools to be considered viable production approaches for the future. This paper will discuss the features and benefits of Hitachi's E-beam lithograph systems and will demonstrate performance by reference to application work from a number of references as well as from Hitachi's own internal use of the tools. Attention will be given to many of the key issues facing present and future lithography, including the manufacture of X-ray masks, reticles and direct write of advanced DRAM devices, proximity effects and new technologies in mask making including phase shifted masks.