This PDF file contains the front matter associated with SPIE Proceedings Volume 12810, including the Title Page and Table of Contents.

Good evening ladies and gentlemen. It's a pleasure being here to speak to you about the IC industry, your major customer. You are a broad group in the international and organizational sense, coming from over 90 companies and 14 countries, but you are also a specialized industry focused on a product that is utilized by another specialized industry--the semiconductor industry. You are steeped in the business and technology aspects of both industries. However, I would like to focus on the business aspects of the semiconductor industry, especially from an international viewpoint.

This was a panel discussion that took place at Bay Area Chrome Users Society Symposium 1986.

Due to the need for better repeatability in mask registration measurement, Intel Corporation instigated a joint project with Nikon Precision Inc attempting to improve the performance of their new XY-2I to +/- 0.05μ over lOOmn. The empirical experimentation focused on controlling and/or more accurately monitoring the errors associated with interferometer compensation for air and mask temperature, temperature differences between the X and Y portions of the interferometer, plate temperature non-uniformity, and incorrect plate climatization. Flouroptic temperature probes monitoring plate and X/Y interferometer temperatures showed that installation of curtains and precisely-placed fans, better placement of Nikon air and mask temperature sensors, and proper plate climatization eliminated X/Y interferometer temperature differences, improved plate temperature stability and uniformity, and greatly reduced the air and plate temperature measurement errors. Long-term repeatability and residual-orthogonality measurements both showed large improvements with the range of orthogonality fluctuations ^ +/- 0.14” of arc and long-term repeatability ^ +/- 0.06μ over 100mm. These results indicate that with a few simple and inexpensive modifications, XY-2I performance improvement is possible.

The LMS-2000 Laser Metrology system is a micro-image dimensional analysis system designed to meet the mask and reticle metrology needs of present and future semiconductor IC designs. The need for improved metrology is driven by the continued advance of micro-lithographic fabrication technology toward the half-micron design rules. The resolution, accuracy, and repeatability of the LMS-2000 system are consistent with the mask and reticle metrology requirements for these design rules. In order to achieve the required 30 nanometer accuracy and the 20 nanometer repeatability for metrology, many factors must be taken into account which have been disregarded in existing metrology systems. These factors relate to a variety of error corrections for the laser interferometer used and the careful control of temperature affecting the measurement. The accuracy and repeatability of the LMS-2000 system result from the capabilities built into the optical scanner used to detect line edges; the interferometer controlled stage and associated calibration and correction techniques; and finally the careful treatment of potential temperature effects affecting measurement accuracy and precision. In addition, throughput, automation, and user friendliness are key attributes which have been incorporated into the LMS-2000 dimensional analysis system.

A laser based photolytic CVD Mask Repair System YL454A, capable of quickly and reliably repairing both clear and opaque defects, has been developed. For clear defect repairing, adhesive micrometer size chromium film of sharp-edged arbitrarily-sized rectangles, can be deposited in the opposite manner of conventional laser zapping process.

In late 1985 and into 1986, several users of pelliclized chrome masks reported the appearance of particulate deposits on the masks under the pellicle. They showed up after 10 days to two weeks in one well documented case, and, reportedly, in as little as a few hours, or as much as a year after the mask and pellicle were mated, in others. In the U.S., the particles were first observed during microscopic examination of the masks, but could also be seen with the naked eye when the mask was illuminated with bright specular light. In Japan they were also found by reticle inspection equipment. The deposits were found along the edges of images and on top of the chrome surface, but almost never on the glass substrate, or on the inside surface of the pellicle membrane (Figures 1 and 2). It was noted that the resist type used to make the mask, optical or e-beam, did not have an influence on the occurrence of the problem, and it was not isolated to the use of any particular vendor’s pellicles. In some instances, it did seem that the problem

"Thin chrome" is a major cause for the rejection of photomasks during their fabrication process. This paper will characterize this phenomenon and identify a working solution to the problem. The thinned regions in chrome are characterized by -100 |xm wide isolated depressions which are detected by visual inspection under an optical microscope. Analysis of the chromium surface using AES and EDXA showed no chemical contamination or inclusions which could be responsible for the thinned chrome. Thin chrome is observed more frequently on plates from Vendor one than from Vendor two. Observations of the removal of the chromium from the glass substrate suggest that the stress gradient within the Vendor one chromium is greater than the Vendor two chromium. ESC A revealed two distinct chemical state of nitrogen present within Vendor l’s chromium film. However, using TEM, the grain structure of chromium from both Vendor one and Vendor two was found to be fine and uniform. The occurrence of thin chrome has been associated with the presence of chromic acid in the resist stripping solution. An alternative sulfuric acid: hydrogen peroxide solution can successfully strip e-beam resist without attacking the chromium mask.

ATEQ Corporation has developed a scanning laser lithography system for use in "state-of-the-art” 5X reticle manufacturing. This system, the C0RE-2000, exposes optical photoresist and can write a typical reticle in less than 20 minutes. Its lithography performance is comparable to presently used electron beam systems but offers higher throughputs and lower costs. This paper presents the results of a detailed evaluation of the CORE-2000 including registration performance, critical dimension control, and throughput. Registration performance and critical dimension control were measured by means of a Nikon 21 Lampas system. System throughput was measured by writing a variety of device patterns, and was compared to the throughput of existing lithography systems.

The objective of this paper is to present an analysis of the use of A21350J as an electron beam resist and the advantages this process technology offers the mask maker in the fabrication of chrome photomasks. The immediate advantage of using A71350J is that it makes use of current processing technologies. The processing of diazo-type resists can be done in two simple, steps: develop and etch. This is a predictable process with repeatable results that does not require any develop cycle endpoint determination. The elimination of the endpoint determination has further advantages in that it eliminates any source of process induced defects as a results of excess handling. The use of AZ resists with an electron beam exposure system results in improved edge quality that can be reproduced in all feature sizes including submicron linewidths. The AZ resist process is compatible with plasma etching. Additional advantages of an AZ process are the uniform critical dimensions, low chrome defect densities, low pinhole densities, and the reproducibility of results. Submicron linewidths can also he resolved with excellent uniformity and repeatability. The low defect densities result in the minimization of the time required for inspection and repair. Further advantages of an AZ process that will be discussed include its wide application, the need for less capital investment than other processes, and its compatibility with FPA standards without requiring any special handling.

Our industry has been hit hard by foreign competition, and it's clear we have to do something to shape up, and fast! We need data, but somehow the reams and reams of data we're known for just isn't enough anymore. Our usual methods of collecting, crunching and reporting the numbers do not do the job of improving and maintaining quality in production. We must improve quality, reduce cycle time, improve yields. It's a matter of survival. The banner some of our managers have been waving is Statistical Process Control, so it's time we learned about it and gave it a try. Statistical process control charts lend themselves quite readily to factory situations, where product may be sampled and measurements and means of the same variable plotted for each data point. A photomask shop doesn't quite work that way. Typically, the number of parts is too low for a sampling scheme to be appropriate, especially in an ebeam shop. Every part is unique, every part is measured for CD's, defects, etc., not just a sample part, and the CD required is different for every part. This paper provides a brief descriptive overview of Statistical Process Control and details the procedures appropriate for a photomask operation. The information and examples are given such that someone with little or no background in statistics may implement SPC procedures in his own mask shop, for the purpose of product quality definition and improvement.

In the late 1970's and early 1980's it became apparent that soft-contact and projection aligners were not going to produce acceptable yields for the one- to three-micron technologies that would be demanded by the mid 1980's. Projection step-and-repeat systems appeared to be the preferred method of wafer imaging, with non-optically generated (e-beam) reticles providing the best masks. Compounded by denser, more complicated circuit designs, this new wave of technology demanded new philosophies for data preparation. With mean-time-between-failure on optical pattern generators shorter than the time it would take to generate some of these new masks, it also became obvious that pattern generators were not going to handle any of the newer technologies. Mask shops were soon faced with the problems associated with multiple tooling techniques, the different data preparation techniques required to support each, as well as the responsibility of maintaining the existing tooling on the older systems.

The majority of VLSI exposure tools used in production are optical systems that use UV light. Optical lithograpy is capable of resolution below 1 urn. Reduced feature size and increased circuit complexity in modern VLSI devices, therefore, impose stringent requirements on photomasks. Photomasks used in wafer steppers and high resolution IX full-field projection systems require high performance such as good chemical durability, pattern stability during cleaning, and high resolution. Chromium films used for conventional photomasks do not adhere well to quartz substrates, resulting in pattern defects due to exfoliation. Furthermore, the fabrication of a Cr mask with high resolution patterns using dry etching is difficult. Pattern defects, caused by poor adhesion, can be generated during photomask cleaning by ultrasonic methods or scrubbing. Dry etching of Cr masks has the problem of low throughput and marginal resist protection. To solve these problems, we have developed a new photomask blank Page 2 using a molybdenum sillclde (MoSi) film deposited on the quartz substrates. A MoSi film deposited on a quartz substrate offers major advantages as a high performance photomask material for VLSI fabrication. There Is no missing pattern due to exfoliation after ultrasonic cleaning with frequency of 28 kHz and 300 W of power, and after being scrubbed over 10 cycles with a high pressure water Jet. Reflectivity and optical density of the MoSi film are not affected by acidic chemicals. Moreover, dry etching with CF4+02 can be aone at a rate 50 nanometers per minute for CMS resist, which is more than five times as fast as the etch rate for chromium masks. Dry etching of MoSi films with a PBS resist has been realized in fabrication of subraicron patterns. In addition a photomask with a MoSi Is easier to repair by PIB than one with chromium because of the higher sputter yields for a MoSi.