Minifabs lower barriers to 300mm

01/01/1999

Minifabs lower barriers to 300 mm

Since 1994, the international community of chipmakers and equipment suppliers has been working toward the introduction of next-generation, 300-mm wafer processing. As a result of the collaboration, global standards were developed and potential conflicts resolved, paving the way for the industry`s 300-mm transition. The recent industry downturn, however, has cooled enthusiasm and delayed new fab decisions, casting doubt on the timing of the transition from 200- to 300-mm processing.

Wafer-size transitions in the high-volume semiconductor manufacturing industry are entirely cost driven. While linewidth reductions can improve device performance and hence average selling prices, the productivity advantage of a wafer-size increase is derived almost entirely from an increased economy of scale. Widespread industry cost models - and the assumptions behind them - may need to be re-examined.

The global semiconductor equipment industry invested billions of dollars to develop 300-mm tools before the industry slowed down. The relatively low-throughput fab, or "minifab," may offer a way to get some return on these investments sooner. The central argument in favor of the 300-mm wafer minifab revolves around chip output.

Today, a high-volume, 200-mm fab typically runs with a throughput of 20,000-30,000 wafers/month. A 300-mm fab would only have to process roughly 10,000 wafers/month to produce the same number of chips as a 24,000-wafer, 200-mm fab. This overwhelming increase in chip output with the larger wafer size has led minifab proponents to predict that most 300-mm fabs would not need to run at throughput levels higher than 10,000 wafers/month. Due to the present DRAM oversupply and fab overcapacity, few people currently advocate super-high-throughput, 300-mm fabs.

Semiconductor foundries have grown in recent years to become large equipment users. Indeed, foundries as a group were a big factor in the equipment industry`s last boom. As OEM manufacturers, foundries do not have their own brand-name products to produce in huge quantities. Rather, they deal with their customers` varying production orders, with device-type and lot-size requirements changing all the time. A smaller fabrication line can be managed more readily to accommodate shorter production runs and minimize work-in-process risks with 300-mm wafers. This production flexibility is particularly valuable for foundries.

Equipment implications

If the minifab idea is widely accepted among chipmakers, equipment suppliers will have to think differently about customer requirements. As fabs are designed to run at lower overall throughput, equipment selection criteria may significantly change. The most profound change would be a de-emphasis on high tool throughput.

Since the early days of the industry, equipment manufacturers have been conditioned by the market to think that higher throughput is always better. Since throughput is a major factor in determining equipment cost of ownership, it translates into cost of ICs produced. While the economics of IC production remain the same, the minifab will cause important shifts in the models for optimum tool design.

For example, assuming a target throughput of 10,000 wafers/month/process and 500 processing hr/month for a well-run operation, then just a single process tool with a modest 20 wafers/hr is all that would be needed in each process step in the fab.

A 300-mm tool with 20 wafers/hr capacity could thus be productive in a minifab environment. Rather than pursuing high throughput as the undisputed top priority, equipment designers should elevate the importance of modularity. To date, modularity has been more of a means to an end for improved, single-wafer process control, as opposed to an end in itself. With minifabs, however, modular tools would be absolutely imperative. The modular approach also gives customers more choice and flexibility in continually optimizing unit processes.

With lower-throughput tools, minifab proponents expect to reap the benefits of smaller footprint and lower cost. However, equipment size and cost are not simply proportional to throughput, as certain basic system functions and components require minimum fixed space and cost to build. Thus, half the throughput does not automatically mean the equipment takes up half the floor space and sells for half the price. For a majority of tools, however, footprint and cost can be reduced as throughput comes down.

Many single-wafer cluster tools today consist of multiple chambers that perform the same process. Thus, another likely benefit of lower-throughput modular tools would be the ability to link different unit-process modules in the same cluster tool. Such physical process integration would reduce overall floor space requirements, and could improve yields through reductions in handling and cycle time (assuming cross-contamination issues can be resolved). Sharing of common platforms would further reduce the aggregate tool cost for a new fab.

Economic implications

A fab today costs about $1.5 billion, with cleanroom and equipment among the largest cost components. If a 300-mm fab is proven to be productive at lower throughput and reduced cleanroom floor space, and if the tools are smaller and cheaper, it will make a compelling case for the minifab.

When fab costs come down, more companies will be able to afford the investment. Thus, it is conceivable more minifabrication lines will be started when the industry resumes its growth, and more 300-mm tools will be sold to outfit them. Chipmakers would be able to enjoy the economic benefits of the larger wafer sooner.

There is a surprise bonus afforded by low-throughput requirements. I still remember some of the frustrations we felt in the early 1980s as we worked to bring up the throughput level in the first Lam Research etcher. The process was good and reproducible, but the initial throughput was not considered high enough. In fact, the effort to increase throughput significantly delayed the commercial shipment of the tool. It comes as no surprise that many new processes have low throughput at the outset, as initial R&D efforts are usually focused on process performance. If throughput requirements are relaxed for 300-mm minifabs, however, new process technologies will be adopted and put to use on the fab floor earlier - to the benefit of all.

Hitachi, Matsushita, Mitsubishi, Sony, and Toshiba are reportedly taking the minifab concept seriously. Quantitative analyses and feasibility experiments are being planned. Their findings and an open debate on the concept will have far-reaching implications for our industry.

David K. Lam, founder of Lam Research, is chairman of the David Lam Group, a management consulting firm for high-growth technology companies, 2055 Gateway Place, Suite 400, San Jose, CA 95110; ph 408/467-3839; fax 408/

526-1588; e-mail [email protected].