This PDF file contains the front matter associated with SPIE Proceedings Volume 12809, including the Title Page and Table of Contents.

This is the keynote address from the Bay Area Chrome Users Society Symposium 1985.

Why bother with mask repair at all? It is all potential to a problem just looking for a place to happen. It comes at the tail end of the process and you don't want to do more damage than you already have done. These two slogans everybody should know: "Make it right the first time" and "KISS", which stands for "keep it simple, stupid." If you think about that for awhile, mask repair violates both of them. If we did it right the first time we wouldn't have to repair at all, and any process that requires an investment of somewhere between $200,000 and $1,000,000 and takes up about ten square meters of clean room space, can't be called simple. Mask fabrication is not as complex as I see wafer fabrication by any means, but the thing is that you still add significant value to a substrate by the time you get down to the end of the process. You start out with something that might be worth a $100-$500 when you buy it—a resist coated chrome plate—and by the time you're finished with you're working on something that's going to be worth $1,000-$4,000. New, if you take a 5X reticle and you had one spot on it you couldn't use it. You have a choice, you're either going to have to trash it or you can go ahead and accept it. If you had to trash it you'd have to repair it, and you think twice before throwing that kind of money away. If you take a look at time, time is important. Any mask maker in the audience will tell that if he has a queue at his E-beam machines, a due-date of yesterday on a mask, mask repair sounds like a pretty good thing to have in your back pocket.

The DRS II System was developed in response to continuous pressure from the semiconductor industry for an improved clear repair system for photomasks. We looked at a number of alternatives to the multi-step liftoff process before deciding on the laser-initiated decomposition of organometallic gases. The desire was to retain the technology of excess chrome removal using a laser zapper, proven in years of use at most mask makers and user installations, with an equally reliable method of missing chrome replacement.

The advent of wafer steppers has given rise to the need for defect-free masks. This requirement presents different problems to the manufacturers of the different types of masks which are used in modern lithography tools. 5X reticles can contain field areas up to 100 square cm which must be free of any defects larger than 1.5 microns. In the case of IX reticles, several areas as large as 1 sq cm must have all defects larger than .5 micron removed. In these two cases, the presence of a single unrepairable clear or opaque defect can cause the rejection of a mask worth several thousand dollars. The ability to make a truly defect free IX full field projection mask would be of great value to the mask user and the mask maker.

This paper briefly describes a new automatic inspection process for 5X reticles with pellicles attached.

Consider the process engineer who has more microscopes than people, the people don't like the scopes, and they are missing too many repeated defects in masks and defects in reticles. These instruments can be made more effective by simple techniques which reduce eyestrain, minimize distracting optical anomalies, and actually make defects easier to see.

At this time I would like to thank the BACUS Association for giving me the opportunity to make this presentation. My talk today is on chrome adhesion on quartz substrate. Since 1985 usage of quartz substrates in our industry has increased as most of us predicted during the last BACUS Symposium. Introduced along with increased usage o-f quartz material, was the difficulty that most vendors encountered in providing a high-quality quartz blank. The difficulty was the insurance of a good chrome adhesion on quartz substrates. Today I will briefly explain how the chrome-to-quartz enterprise condition is different from that of chrome to soda lime and chrome to RE, and what Ulcoat has done to insure adhesion qualities that is being evaluated by the users of quartz material.

One of the continuing issues in the manufacturing of IX reticles is defects. Due to the nature of the 1 — to—1 Stepper, defects do play a large role in the production of IX reticles. One of the problems with defects is the ability to define what role a particular defect will play when manufacturing wafers. To understand exactly what effect defects have on a 1-to-l Stepper, Ultratech has started an in-depth evaluation on' defect printabi1ity. The goals are to: 1) Define what a printable defect is. 2) Define the effect of a particular defect type. 3) Define what role a particular defect has on a particular geometry. 4) Aid Ultratech Stepper users in understanding what defects, and/or location of defects, to concentrate on evaluating. 5) Assist mask makers to optimize their inspection to find those defects which would be destructive. 6) Inform equipment manufacturers, specifically inspection and repair, what needs will be expected for present and future generations of IX reticles. In this ongoing study, defect printabi1ity has been researched extensively. The results gathered to date are what shall be discussed in this paper.

During this presentation, I would like to introduce the concept our H-series blanks which we have developed in conjunction with our new production line. I'd like to present a comparison between quality of our former blanks and the new HQZ material now offered. Finally, I want to address the effect of defect on the substrate which are not presently detectable using automated inspection equipment after the final manufacturing phase.

The standard mask making process involves 1) resist patterning with E-beam or optical lithography and (2) pattern transfer to chrome film. The quality of the resist layers (defects, uniformity, etc.) frequently determines the quality of chrome blanks. Furthermore, the quality of resist pattern (defects, critical dimensions, etc.) frequently determines the quality of final masks. Therefore, a new technique is required which will allow manufacturers to examine the resist layer itself. Fluorescence microscopy, using either intrinsic light emission from resist material or induced emission with additives (dyes), will satisfy these demands. E-beam resists have been incorporated with fluorescence dyes and successfully exposed and patterned with standard production procedures. The enhanced contrast resulting from fluorescence techniques allows (1) easy identification of micro-flaws and contaminations, (2) accurate and precise measurement of CDs and (3) close monitoring of end-points in developing, de-scumming and stripping. Parallel experiments have also been done with optical resists. Care must be taken, though, to avoid premature sensitization of optical resists. A careful evaluation of this technique in a production environment (blank manufacturing and mask making) will be presented. Possibilities of resist image repair prior to pattern transfer will also be explored. A new class of low defect blanks will also be discussed. These blanks were produced in a highly-automated . production line, with in-line sputtering machine and hot-plate bake. Extensive evaluation at MMI over a two month period indicated that the state-of-art material can support at quality level of 0.5 DPSI (prior to repair) at 68.67. as compared to that of the standard product at 38.51%. Defect data from KLA included all pinholes, opaque spots, and perm at 1.0 - 1.5 micron level. Comparisons between plate quality of different vendors, both domestic and foreign, will also be presented.

Currently, the most popular method among various types of photolithography the step and repeat camera. Linewidth control using this type of lithography is one of the key elements for both mask maker and reference publishers. Considering all of the parameters and the mechanism to determine the critical dimensions, is very complicated. At Toppan we have been attempting to estimate the amount critical dimension movement mathematically and experimentally prior to final publication.

The most important step in the processing of PBS electron beam resist is the development step. It has the greatest influence on the quality and appearance of the resist image. The purpose of this paper is to study the effect of developer concentration on PBS development/dissolution rate and resist image using optimized PBS resist coated mask blanks. The critical dimensions will be measured to determine the development rate and the image quality will be evaluated empirically. We'll be holding all other process variables constant.

There are many commercial positive type electron beam resists available for mask making. The most commonly used resist in the United States is the PBS type. U.S. mask makers seem to prefer this type because of its high sensitivity and high throughput in regard to E-Beam writing time. However, there are difficulties and instabilities exhibited by PBS during the process cycle which cause major concerns. Mask makers in Japan use EBR-9 as their primary positive E-Beam resist for production of reticles, masters, and sub-micron masks. With regard to processing and other factors, such as shelf life, adhesion, defect density, etc., EBR-9 has proven much more stable than PBS. The major concern with EBR-9 has been its lower sensitivity (about 5uc/cm sq.), which reduces the electron-beam system's throughput. This paper will present an evaluation of EBR-9 resist for mask production and techniques which give EBR-9 sensitivity and throughput comparable to PBS.

One o-f the key -features in the development of X-ray lithography as a viable alternative to optical techniques is the technology of X-ray masks. Shifting from optical to X-ray sources requires significant changes in maskmaking techniques. Glass substrates can no longer be used, but must be replaced by thin membrane structures. The membrane is X-ray transparent and is used to support a metallic absorber pattern. Boron nitride is the most commonly used membrane material to date. Chrome is not an adequate absorber and must be replaced by higher atomic number materials, such as gold. Much thicker absorber patterns, in the range of 0.5 to 1.0 microns are required. Since all X-ray lithography systems are IX proximity printers, processes for absorber patterning must have resolution in the submicron range. Two techniques for gold absorber patterning have been developed and are currently in use at Micronix Corporation. Subtractive patterning is accomplished by sputter etching the gold absorber layer using a tantalum etch mask. This technique produces sloped sidewalls, which limit the resolution achievable. An improved process, using gold electroplating into a resist stencil, has been developed. This additive technique has been shown to produce absorber patterns with vertical sidewalls for dimensions as small as 0.5 microns.

The first reports of this phenomena that I found was from about I960. In 1978 in one of the references that you'll find in the printed version of this, are some patents on some amine containing salts. These salts, when doped into positive resist, produce image reversal from positive to negative when it was exposed. In 1982 a paper given by a group at IBM at one of the Kodak Interface Symposiums talked about this in detail, trying to demonstrate what they thought the mechanism was of how this miraculous phenomenon worked.

I am grateful to the BACUS organization for this opportunity to talk about the use of a personal computer tor process control. We are living in an information society and this paper deals with getting the right information at the right time. Having been educated as an engineer, I still believe that for every problem there is a solution — so long as we have all the data we need. In a mask shop, or for that matter, any operation, the data we need to improve the operation efficiency is right there before our eyes. The difference between companies that do improve and the ones that do not is usually the accessibility of information and the application of the information received. Keeping this point in mind, I would like to now discuss the use of a personal computer for process control and for generating greater revenues. At Master Images, we have been using a personal computer for tracking critical dimensions, registration and defect density by customer, product, exposure system, material, process, mask field, time, inspection systems and operators involved with producing the photomasks. All data are stored in a Symphony data base with data validation check. All calculations are done automatically on Symphony spreadsheet. The computer we use is an IBM PC. An operator with no previous computer knowledge can be trained in approximately one hour to enter data and generate a comprehensive twelve-page quality report. Today I have brought with me a few slides that show some of the benefits of using a personal computer to track photomask quality.

Last night Steve Dunbrach and the panel discussed the problem of ever tightening specifications and equipment limits. Do the mask makers have their facilities and procedures optimized to push the equipment to the limits? Answers to these questions may make some opportunities apparent: 1. How close is your turn time to the equipment limits? Suppose that you are given one month advance warning on the design sign off date for a megabit chip at a remote site. How long would it take your shop to complete the first mask in the set? The typical answer is 3 to 5 days. 2. How many hours would that first completed mask take in one of the most advanced shops in the U.S.? The answer at AT&T is 8 hours. 3. How clean is your QA department during production? A typical answer is Class 1000. 4. How clean is the GW department in one of the most advanced shops in the U.S.? The answer at AT&T is Class 10.

The first slide shows what happens to lithography when you expose through a pellicle that's been contaminated. This particular pellicle was contaminated by a finger print and no one would be happy with that kind of distortion and side walls. This is on a wafer, by the way, and I think a mask would similar. You can really see the distortion at 1000X total magnification. Essentially, there are two types of contamination problems relating to pellicles. Obviously, most particles that we find in a typical mask shop or wafer fab are not going to really be a problem. Anything sixty microns and below is not going to bother lithography on most any kind of equipment you use. However, fingerprints from handling, operator, and can moisture from present problems either the nitrogen lines or from the operator, can present problems. At SSEC we have been working for quite some time developing cleaning techniques for these types of pellicles. There have been some things written in the industry saying that you can't expose pellicles to water or water-based detergents because it creates problems. However, we have devised a technique using our mask cleaning equipment that is consistently reliable. The process that we have developed is a very simple detergent cleaning using an oscillating low-pressure dispense arm, that can be programmed back and forth over the plate surface while it's spinning. The detergent rinse is Microclean, followed by water rinse, which is followed by a very high speed 5000 rpm spin dry. One of the key elements in this cleaning process is the ability to oscillate the dispense arm over the plate, making contact with the detergent over each part of the pellicle, rather than dispense in the center of the plate and allowing the centrifugal force, caused by the plate's spinning, to just glaze over the top of the surface. That tends to leave some film on the plate.

One of the reasons we are gathered here this evening is to honor one of our own. The directors of BACUS felt that we are a growing, dynamic, important industry unto ourselves, and that peer recognition of achievement was long overdue. To correct the situation we have instituted what we hope to be an annual event — the awarding of the BACUS Prize, to be given to the person or company that has made an outstanding contribution to the photomasking industry. The prize consists of a personal plaque and a cash award of $500, to be give to the educational institution of the recipient's choice. This year's recipient is a person who has changed the way we do things for the better. Just a few short, but memorable, years ago, plate cleaning was done with Ivory soap and engravers' cotton. Plate processing was done by the "dunk and look" method until the process, in the estimation of the operator, was completed. Jim Giffin changed all that for us.

These are the closing remarks from the Bay Area Chrome Users Society Symposium 1985.

Panel Discussion from Bay Area Chrome Users Society Symposium 1985.

The manufacture of reticles for the Ultratech Model 900 IX wafer stepper presents the greatest challenge in mask-making. Each finished reticle must meet the strictest standards and specifications in defect density, critical dimension control and registration and overlay accuracy. The IX reticle places great demands on the ability of the mask maker to produce fine line lithography, typically 1.5 microns and smaller, with no defects larger than 0.5 microns. The effect of these manufacturing requirements on the inspection process is reviewed in this paper. What is the best manufacturing flow to assure economical use of very expensive capital equipment needed to inspect and qualify this product? Decision criteria for resolving these issues are set forth.