In order to be able to write 64 megabit DRAM reticles, to prepare to write 256 megabit DRAM reticles and in general to meet the current and next generation mask and reticle quality requirements, Hoya Micro Mask (HMM) installed in 1991 the first CORE-2564 Laser Reticle Writer from Etec Systems, Inc. The system was delivered as a CORE-2500XP and was subsequently upgraded to a 2564. The CORE (Custom Optical Reticle Engraver) system produces photomasks with an exposure strategy similar to that employed by an electron beam system, but it uses a laser beam to deliver the photoresist exposure energy. Since then the 2564 has been tested by Etec's standard Acceptance Test Procedure and by several supplementary HMM techniques to insure performance to all the Etec advertised specifications and certain additional HMM requirements that were more demanding and/or more thorough than the advertised specifications. The primary purpose of the HMM tests was to more closely duplicate mask usage. The performance aspects covered by the tests include registration accuracy and repeatability; linewidth accuracy, uniformity and linearity; stripe butting; stripe and scan linearity; edge quality; system cleanliness; minimum geometry resolution; minimum address size and plate loading accuracy and repeatability.

The semiconductor industry has directed its course toward volume production of 64M DRAMs. Feature sizes of typical 64M DRAM are 0.3 to 0.4 micrometers , and the linewidth and positioning accuracies required for 5X reticle production are both 0.05 micrometers . To respond to these requirements, JEOL has developed an electron beam lithography system for reticle making, JBX-7000MV, which incorporates improved variable shaped beam optics to assure high speed and highly accurate pattern writing. The high accuracy writing was made possible by: (1) reducing the pattern data increment and correction increments by 1/2; (2) applying a field shift writing method; and (3) a new substrate holder insensitive to the weight of a mask. This paper introduces the JBX-7000MV and its capabilities, and discusses the issues required for second generation 64M DRAM reticles. Anticipated improvements expected to be integrated in the JBX-7000MV to meet these requirements are discussed.

Performance specifications, electron gun parameters, testing results to date, and throughput are presented for the second prototype EBES4 which is entering manufacture. The thermal field emitter electron gun is designed for high flux, high current stability, and a long lifetime. The gun produces 250 nA into a 125 nm diameter spot for a flux of 2000 A/cm² with a beam current drift of <= 0.5%/hour. Minimum address size for EBES4 files is 1/64 micrometers (16 nm). Measurements of butting and shear errors for the 256 micrometers stripe and 32 micrometers subfield boundaries indicated an accuracy (mean + 3(sigma) ) of <= 60 nm without correcting for reading precision. Image placement accuracy (mean + 3(sigma) ) of <= 50 nm was measured on an LMS 2000 for MARKET cross arrays spanning 100 mm X 100 mm. Overlay accuracy (mean + 3 (sigma) ) for multi-point alignment of three masks was measured as <= 50 nm with a maximum error of <= 30 nm.

Even at 0.5 micrometers design rules, the specifications on 5X reticles are set extremely tightly, on the grounds that ULSI patterns are so complex that tight specifications are essential to obtain acceptable yield. It is usually assumed that these specifications scale with the design rules, and that they should be even tighter for 1X reticles. As a consequence, it has been argued that 1X reticles for 0.25 micrometers design rules are impracticable. A statistical analysis, starting from first principles, and assuming point independent, normally distributed errors, supports the way in which mask specifications are currently set. The assumptions of spatial invariance and normal distribution are crucially important in the analysis. However, it is far from clear that they are valid. Consequently, mask specifications in general, as they are currently set, may be unnecessarily severe.

Two lithographic processes for phase-shift mask (PSM) manufacturing have been investigated. In particular, processes in E-beam (electron beam) lithography by use of a charge-dissipating layer of a conductive polymer are studied. Two commercial conductive polymers, TQV and ESPACER100, are found to work well for charge-dissipation. Three new resists along with CMS and EBR9 are evaluated regarding their properties necessary for patterning a shifter layer. Among them two new resists are demonstrated to be excellent. The effect of the number of data-blocks on the alignment accuracy is examined in delineation with a Hitachi HL-600, where each data-block has four fine-alignment marks. The examination suggests that the use of one or two data-blocks is practical. As to combination of writers for the Cr level and the shifter level, HL-600 - HL-600 gave better alignment accuracy than the other combinations, WW6000 - HL-600 and MEBES III - HL-600, did. The comparison between the E-beam and the laser writers is summarized.

As mask CD precision and uniformity specifications continue to tighten to 40 nm and below, a better understanding of CD error contributors is required in order to achieve this level of performance. The lithography system, the resist and chrome coatings, development and etching, and metrology all can contribute to the composite CD error. Error sources as small as 10 nm can be significant contributors depending on how the error adds in to the total. In this paper, techniques for isolating and examining these errors are discussed. The CD performance of the CORE-2564 is improved through the systematic optimization of process steps. The implications of `zero-bias' processing are discussed.

An improvement in the method used to fabricate 5 X 5 in² -1 and 5x biased and unbiased optical masks is achieved by actively controlling the resist development step of the mask fabrication process. This method has been initially applied to a photomask process which utilizes poly(1-butene-co-sulfur dioxide)(PBS) resist as the pattern delineation material. Real-time targeting of resist feature dimensions is performed using a Laserlith Resist Thickness and Endpoint Controller which has been adapted to an Applied Process Technology^R/Convac Model 915 resist processor. The controller monitors in real-time the one-step resist development process for a time period based on the measured development rate of the resist, the geometry and size of the targeted feature. After the resist is developed, the controller instructs the resist processor to continue onto the remaining steps in the processing cycle. The targeted resist features of initial product produced using this system have an average variation from targeted size of 0.04 micrometers and an average resist linewidth uniformity (3(sigma) ) of 0.04 micrometers . These results indicate that active control of this critical development step enables the resist feature dimensions to be within +/- 0.05 micrometers of their targeted size after completion of the post-development bake step.

The resolution and critical dimension control requirements for photomask fabrication are increasing at a dramatic rate due to advances in wafer lithography systems and photoresist technology. For example, phase shifting techniques for 5x reduction steppers require subresolution phase shifter elements as small as 0.5 micrometers to be patterned on the reticle. Unity magnification systems such as 1x optical steppers and deep UV 1x steppers require sub- half micron resolution on the reticle. The latest generation of electron-beam mask making systems is capable of patterning these structures in the resist film. However, traditional wet etch is not capable of successfully transferring the pattern from the resist into the chrome. This paper discusses a dry etch chrome process that has been developed at TRW. Sub-half micron resolution is characterized and explained in terms of chrome etching parameters. Selectivity and process sensitivities are explored for a potential manufacturing process. Finally, a dry etch process is used to fabricate actual reticles for an Ultratech 1500 1x optical stepper for use in a wafer manufacturing line.

The forecasted specifications for 64 meg 5X reticles dramatically reduce the allowable range for CD accuracy and imprecision. These 1992/93 5X reticle CD requirements will demand not only that the photomask supplier manufacture reticles to uniformity requirements unheard of only last year, but also measure them with enough precision to avoid yield loss caused by measurement imprecision. Of fundamental importance is how yield loss is caused and affected by measurement sample size, process capability, measurement specifications, and the magnitude of measurement tool imprecision. This paper describes a cost model developed for metrology tools. The model uses standard formulations of Cp, precision to tolerance ratios (P/T), and measurement tool imprecision to forecast yield loss based on measurement sample size and reticle specifications. The model gives a close look at Cp distortion caused by measurement tool imprecision and what Cp distortion's impact on manufacturing cost is. A `typical' manufacturing environment is modeled and a cost analysis done.

This paper addresses the various issues of critical dimension (CD) measurement methodology for mask manufacturing today, and suggests a new approach for standardization of CD process monitors, process improvement/development, and process capability calculations. The sampling method -- both number and location -- used to measure CD performance has become more crucial to high density, fine line device fabrication. Unfortunately, most of the mask manufacturing industry still use either CD cells or a few special CDs located outside the device area to define how well the reticle's CDs meet the specifications. As an alternative, for process monitoring, process improvement/development and process capability calculations, we suggest using an `artifact,' a specially designed test mask. This mask allows measurement of a large number of CDs in order to accurately calculate the mask CD control. Using an `artifact,' process engineers can characterize CD patterns in their processes and can make better informed decisions. Unusual patterns in the CDs are more easily recognized with artifact data and, therefore, can be analyzed more efficiently. By measuring process capability with artifact CD data, rather than with product CD date, customers and management can be provided with a more accurate measure of the CD performance of a mask manufacturing line. In this paper, the application of artifact CD data to measuring the capability of a process to meet a CD performance specification is studied in detail.

Coherence probe metrology (CPM) is a unique optical imaging technology which allows non- contact three dimensional measurements of sub-micron features. This technology has been applied to CD and overlay metrology on semiconductor wafers. CPM technology employs a Linnik interferometer to collect 3-D information, measuring both the amplitude and phase of an image. This makes CPM a promising technology for phase shift mask metrology. This paper gives a technical description covering the theory of the coherence probe microscope. Cross-sectional CPM images of phase shift mask features are presented, showing the ability to image quartz and PMMA shifter structures. Initial metrology performances for feature width and shifter thickness are presented. Linearity of CD is examined. Results are shown for shifter only and rim shifter masks. Shifter thickness measurement precision is compared for envelope-only and phase based signal measurement algorithms.

The etch depth of phase shift masks is typically measured by means of profilometry and the expected phase shift is calculated from a knowledge of the refractive index at the lithographic wavelength of interest. In the case of masks utilizing deposited films the index may differ from values for bulk materials and commonly varies to some degree with method of deposition. The interferometer offers a method for the measurement of phase directly and, hence, for a means to obtain refractive index values for phase shift films on quartz blank substrates. Arrangements are described for direct phase measurements using visible, 632.8 nm, and UV, 257 nm, radiation.

The pelliclization process at Intel during the first half of 1991 was not in control. Weekly process yield was trending downward, and the range of the weekly yield during that time frame was greater than 40%. A focused effort in process yield improvement, that started in the second half of 1991 and continued through 1992, brought process yield up an average of 20%, and reduced the range of the process yield to 20 - 25%. This paper discusses the continuous process improvement guidelines that are being followed to reduce variations/defects in the pelliclization process. Teamwork tools, such as Pareto charts, fishbone diagrams, and simple experiments, prioritize efforts and help find the root cause of the defects. Best known methods (BKM), monitors, PMs, and excursion control aid in the elimination and prevention of defects. Monitoring progress and repeating the whole procedure are the final two guidelines. The benefits from the use of the continuous process improvement guidelines and tools can be seen in examples of the actions, impacts, and results for the last half of 1991 and the first half of 1992.

Significant effort is expended at photomask facilities to achieve particle-free reticles prior to pelliclization. The smaller the specified particle size, the more difficult this task becomes. Laser scanning equipment is now available which can locate particles in the pellicle cavity down to a specified 0.5 micrometers . Some optical techniques are more sensitive but less repeatable. Pixel scanning systems can find even smaller particles. In this study two test reticles, one with programmed defects in relatively large geometries and one with programmed defects in highly periodic structures were contaminated with poly styrene latex (PSL) spheres in sizes from 0.5 micrometers to 4 micrometers . The reticles were characterized and printed onto wafers to determine the printability of the spheres compared to programmed chrome defects with the same size and situation. The smallest PSL sphere which printed was 0.7 micrometers and it occurred in periodic structures of 0.8 micrometers .

The use of post-pelliclization inspection is a step toward zero customer return, due to particles on the mask under the pellicle. Implementation of a laser inspection system in wafer fabs allows mask users with the ability to (1) check the quality of incoming masks for shipping related defects, (2) perform on-going reliability checks to reduce mask defect related yield losses, and (3) reduce handling by reducing manual inspection. This paper describes the techniques used in the calibration of the laser inspection system, QC Optics API-3000, and several benefits from this inspection. In addition to calibration, monitoring system performance is critical to the success of this inspection. Types of monitor masks are described here, including methods of generation and uses of monitor masks.

Phase shifting masks (PSMs) have shown great promise for extending resolution in optical lithography. However, the production of defect free PSMs is a challenging task. Because of their increased sensitivity to process faults and defects, production difficulty is a major hurdle to overcome before PSMs become commonplace in semiconductor manufacturing. Rigorous simulation of PSM defects and production faults is a useful tool for understanding the sensitivities and limitations of this new technology. The complex topographies and non- planarities of PSMs require vector modeling to properly consider light scattering effects. Scalar models such as SPLAT cannot investigate scattering due to high edges, thick layers, and changes in refractive index. We are investigating PSMs using the rigorous 2-D photolithography simulator METROPOLE. In particular, we are studying the advantages and limitations of two new techniques for PSM defect repair. The first technique uses a double shifting structure, the second uses an electron beam sensitive silicon-containing resist.

In the past several months attention has been focused on improving the chrome repair technology, an equally important consideration for future photomask applications. In this paper a new Nd:YAG laser configuration is described. Chrome machining characteristics of the Nd:YAG laser are discussed in terms of the cut-to-cut dimensional stability and the substrate surface quality after repair. Semi-automated stage motion and positioning techniques have been developed to take full advantage of the improved reproducibility of the Nd:YAG laser. It is shown that repair quality has been significantly improved over the full range of repair dimensions compared with the Nd:YAG laser configuration currently employed in laser mask repair tools. In conclusion, the implications of these improvements for future mask applications, including phase-shifting masks, are discussed.

In the past, database inspection of photo masks has been used primarily as design verification; defect inspection was done only as a last resort. The increase of single die reticles has forced increasing use of database for defect inspection, but at the cost of sensitivity and through put time (TPT). With the advent of fast throughput and high sensitivity database systems, the use of database for inspection of photo masks is taking on a new role. The comparison of chrome on glass photo masks to the ideal database can now be used to quantify mask parameters such as process bias uniformity, edge roughness, and transmission uniformity in addition to defects. Database inspection will move from the realm of a last resort option to a preferred option. The user must now be prepared to understand and use the valuable data now available to them to quantify the quality of their mask, and improve their process. Intel designed artifact masks were created to quantify process bias uniformity in addition to other defects. Using mask to database inspection, critical dimension (CD) process uniformity, and edge roughness in addition to both traditional and thoroughness defects, were quantified throughout the active area. Results presented demonstrate the additional information now available to the mask engineer to evaluate mask quality, and implement process changes.

The status of x-ray mask inspection and repair at the IBM Advanced Mask Facility is presented. Defect classification and sources are presented along with some preliminary results from a defect printing study done at the Advanced Lithography Facility.

This paper covers what the Council for Continuous Improvement (CCI) has learned about the most cost effective way of implementing Continuous Improvement and how, in doing so, it will have the greatest impact on your company.

Empowering employees on the firing-line to actively participate in solving business problems can have a significant positive impact on bottom-line performance. Lessons from a number of companies and thousands of people from the firing-line have demonstrated this. Unfortunately, there is no simple, step-by-step method that can be guaranteed to succeed. Getting business results by successfully empowering the firing-line is as much art as science. Yet, as with any art, there are principles that, if rigorously followed, can go a long way to ensuring success. These principles are outlined in this paper as 18 critical success factors to empowerment.

The impact on image quality of scattering from phase-shifter edges and of interactions between phase-shifter and chrome edges is assessed using rigorous electromagnetic simulation. Effects of edge taper in phase-shift masks, spacing between phase-shifter and chrome edges, small outrigger features with a trench phase-shifter, and of the repair of phase defects by etching to 360 degree(s) are considered. Near field distributions and diffraction efficiencies are examined and images are compared with more approximate results from the commonly used Hopkins' theory of imaging.

A new, simple approach is given for the design of phase shifting masks using self-aligned rim shifters. First, a series of design curves are generated, which show all the combinations of rim width and chrome width which print a certain feature for a given process. Second, based on performance criteria for the critical feature(s) on the mask, a rim width is chosen. Finally, the design curves are used with the desired rim width to generate a bias rule which determines the needed chrome mask bias as a function of feature size and type. The result is an easy-to- fabricate self-aligned rim shifter mask with nearly the same performance benefits of a complicated two-level sized rim shifter mask. This technique is called the Biased Self-Aligned Rim Shifter.

Use of the phase-shifting mask in lithography allows substantial improvement of the resolution of commercial steppers for periodic circuit patterns, and, in certain cases, features smaller than 200 nm can be fabricated using i-line steppers ((lambda) equals 365 nm). There has therefore been much interest in developing this technology in recent years. In this paper, we examine and compare several approaches to evaluating mask designs for terminated periodic features, narrow gate lines, and contact holes, and compare the simulations with the actual results obtained when one attempts to use these designs in practice. Although we have found atomic force microscopy (AFM) to be a key tool for metrology, we conclude that there is a vital need for mask simulation, fabrication, inspection, and repair techniques to be developed further before these `imaginary' masks can be useful in the real world.

Laser scatterometry is a noncontact, rapid method of collecting and analyzing light scattered from a structure. We have applied optical scatter techniques to measure the surface roughness as well as the etch depth of phase shifting masks (PSMs). Experimental results and theoretical modeling are discussed.

The fabrication of phase shifting masks requires precise alignment between the primary and shifter layers. The MEBES^R IV electron-beam lithography system uses its SEM mode to acquire a video image of the phase shift mask (PSM) alignment mark. Digital signal- processing algorithms have been developed to accurately determine the locations of the marks. Alignment marks are acquired through various resist systems and film thicknesses. Machine control software translates and rotates the MEBES coordinate system to align it with the mask coordinate system, as determined by the location of the alignment marks. Results showing overlay accuracy between layers are presented.