Maskmaking gains new prominence as device rules shrink
11/01/1997
Maskmaking gains new prominence as device rules shrink
Dave Miller, Etec Systems Inc., Hayward, California
As 1997 nears an end, the pattern generation and maskmaking sector of the semiconductor industry is ramping up to provide the next wave of change, as device design rules obey Gordon Moore`s famous law and prepare to shrink once again. This year we will see 0.25-?m semiconductor devices in volume production with 0.18-?m devices not far behind. Improvements in mask pattern generation (PG) and advanced maskmaking techniques are critical milestones on the industry`s path to smaller design rules.
Good times for maskmakers
The maskmaker faces many challenges in the next few years, and will need to devote significant resources to developing new processes and purchasing new equipment, while gearing up for what looks to be another time of significant growth. Mask sales passed $2 billion in 1996 and, by several estimates, will reach $4 billion in 2001. Maskmaking is a good business to be in today, and indicators point to a strong future, with another fast growth period expected at the end of the decade. It has not always looked so good - the maskmakers` seven-year vacation in the 1980s had a substantial negative impact and caused maskmaking to lose some of its perceived value to the industry (see "Maskmakers` vacation," on p. 70).
 |
Figure 1. SIA Roadmap (initial production). Critical level mask requirements.
 |
Figure 2. The role of mask pattern generation.
Maskmaking and mask PG have, in the last couple of years, gained new prominence as the semiconductor industry realizes that its move to the next generation of devices hinges on the maskmakers` ability to deliver the new, more complicated masks well ahead of full device production requirements. A successful maskmaker must invest in PG equipment and new processes that will enable this move to smaller designs, and must have strategic plans in place to meet the SIA Roadmap`s mandate of 0.13-?m and small devices early in the next century.
As feature sizes shrink, masks are getting much harder to make. Future wafer lithography will use lower magnification than the current 5? standard. 2? and 2.5? for noncritical layers are being implemented in some locations. At these lower magnifications, noncritical layer masks are almost as difficult to produce as critical mask layers. PG machines whose lives had been stretched for three device generations during the vacation are now unable to produce the leading-edge masks required for new devices. This stimulates machine sales.
Further indicators of strong future growth are the complicated nature of new processes and new maskmaking techniques. Due to optical proximity effect correction (OPC), the feature size on the reticle is shrinking faster than the feature size on the wafer. Phase shift masks (PSMs) reduce mask throughput as denser designs and tighter tolerances slow down throughput. Both of these changes require more PG equipment to produce the same number of reticles, raising costs in the maskmaking facility.
Over the last three years, the maskmaking sector has grown significantly. Continued investment in capacity and capability is paying off, as is a modernization of business practices and improved relationships with their customers and equipment suppliers. According to Mike McCarthy, director of corporate communications for Photronics Inc., "It makes a great deal of sense for the captives to outsource to the merchants. We have made the investments in equipment and capacity, and the experience gained from our volume production makes us a good team player. As masks become more complex, there is increasingly less technical benefit maintained by the captives." Outsourcing volume production to merchant shops allows captive maskmakers to increase their focus on leading-edge process development and future device generations.
All major merchant maskmakers have become publicly traded companies, such as Photronics and DuPont Photomask. In the last decade, there has been a shift in the number of masks produced by the merchant vs. captive merchant maskmakers. This trend should continue as manufacturers increasingly rely on the technical capabilities and expertise of the maskmakers and outsource production to them.
Mask equipment markets must grow as well
We are seeing the beginnings of the switch to the next level of devices, with 0.35-?m devices in volume production, 0.25-?m starting to be produced at higher volumes, and pilot production of 0.18-?m having begun. Investment in PG equipment will grow as the maskmaking sector races to meet requirements for the next generation devices. Some market researchers are projecting a doubling of the market for pattern generators as a result.
The SIA Roadmap sets aggressive technology goals, has accelerated technical requirements, and is shortening technology cycles. This has led to an enormous expenditure on R&D in PG technology. Cooperation is becoming more and more critical as R&D costs reach unprecedented levels. Government agencies, such as Sematech and DARPA, came to the rescue of suppliers in the early 1990s, and government funding, strategic alliances, and industry consortiums continue to help defray the high cost of PG R&D.
To be successful, the maskmaking equipment manufacturer must be able to reach the new design rules in time to meet the SIA requirements. Manufacturers must also be prepared for unprecedented R&D costs, which are expected by some analysts to more than double for those seeking to deliver PG solutions for sub-0.18 ?m.
Recently, maskmaking and mask PG have gained new prominence with shrinking device rules (0.35-?m) and a rapid acceleration in worldwide IC production. Devices became more complex, and the average number of process layers/IC design continued to increase. The market for pattern generation equipment is typically countercyclic to devices themselves. At times when IC revenues have fallen, IC manufacturers typically focus on new designs that increase mask demand. Manufacturers have begun to order equipment for 0.18-?m device manufacturing, despite the fact that they will not be producing these devices in volume for several years.
What is the future for maskmaking and mask PG?
Research is under way at Etec Systems to develop PG systems for 100-nm and smaller devices. The challenges facing this, and every step down in device rules, are formidable. For example, mask sizes, data paths, and throughput are challenges that will require unprecedented R&D to produce the value-added features necessary to conquer the challenges. Mask sizes will reach nine inches in the year 2000, and the SIA Roadmap offers the PG suppliers and maskmakers less time to find solutions than before (Fig. 1).
Changes and value-added features should provide the leverage for maskmakers and PG equipment manufacturers to increase prices and drive a more profitable business. Maskmaking revenue, as a percentage of IC revenue, at one time was more than 3%, falling during the 1980s to near 1%. The figure shows that these revenues are rising, as manufacturers accept the value of the mask to their business and masks become less of a commodity.
Of the two current commercial PG technologies-vector scan and raster scan-the latter is the most likely method of choice and at the moment, the only solution below 0.25-?m rules. Most in the industry feel that raster scanning will continue to be the method of choice due to its industry leadership, superior accuracy, and ease of data preparation. Electron-beam direct-write in volume production is many years away.
The 130-nm goal of 2003 is a challenging one. Maskmakers have emerged as a vital sector of the semiconductor industry. "Masks used to be a passive image transfer medium," stated Ken Rygler, executive VP of worldwide marketing and strategic planning for DuPont Photomasks Inc. "For the 1? designs of the 1980s and early 1990s, the mask was basically an exact replica of the image it projected. With today`s OPC and PSM techniques, we are now introducing features that are not printing, but add value to the final product. The mask then becomes a value-added contributor to the lithography cycle. These masks cost more, contribute more, and ultimately are worth more." For maskmaking equipment manufacturers, the challenge is to supply new equipment well in advance of the SIA Roadmap`s production requirements.
In the mid-1980s, a period termed "Maskmakers` vacation" began -the result of optical lithography advances that reduced the need to purchase new PG equipment and reduced overall mask demand. Four factors were the cause:
Wafer fabs quickly adopted 5? technology. Five times larger features on a reticle meant existing maskmaking machines could be used for further generations of ICs.
5? reticles could be written much faster than a 1? mask for the same design. This left existing capacity that was more than enough to fill the demand.
The use of pellicles to cover and protect reticles reduced demand for new reticles.
Electronic design automation reduced test iterations of a design prior to production, thereby reducing the number of masks that could be written.
Overcapacity and flat mask demand held the price of the mask artificially low for many years. A consolidation in the industry began as a few major mask producers merged small shops into larger new companies, such as Photronics and DuPont Photomask. Mask revenues dropped to approximately 1% of semiconductor revenues, from nearly three times that in the 1980s (see figure). Investment in R&D and new equipment reached a historical minimum in 1992.
 |
Dave Miller is marketing communications manager at Etec Systems. Etec Systems Inc., 26460 Corporate Ave., Hayward, A 94545; ph 510/783-9210, fax 510/887-2870.