Feature
Index
Semicon West 2002| thinfilmmfg.com
Feature
Index
Semicon West 2002 |
If a single theme emerged from July's Semicon West meeting, it was that advanced technology is expensive. Basic research, process development and integration, and production ramp all cost lots of money. The 90 nm and smaller technology nodes will require unprecedented investments. Manufacturers and equipment suppliers alike must continue to support research programs even as they face the second year of one of the worst downturns in semiconductor industry history.
As KLA-Tencor's Tom Long explained in a panel discussion on advanced lithography, the 130 nm node brings several major inflection points, ranging from widespread introduction of copper interconnects to deep sub-wavelength lithography. Designers must contend with more difficult signal integrity and leakage issues. Photomask manufacturers and lithographers must incorporate radical resolution enhancement techniques. Process engineers must face all the difficulties inherent in copper interconnects and (at the 90 nm node) low-k dielectrics. Many chip vendors have encountered yield issues during prototyping, Long said, forcing them to adopt less aggressive design rules.
Thus, there is no binary transition to the 130 nm node. As with past transitions, the move to 130 nm will be gradual, with the majority of production trailing years behind the leading edge. The challenge for equipment makers is to continue to serve all segments of the increasingly diverse market.
As I talked to show attendees, an apparent contradiction appeared. Informed opinion agrees that 130 nm designs have been slow to appear and slow to ramp to production. Yet most foundries are reporting high utilization rates for their leading edge capacity, usually a sign that expansion is needed. The foundries, however, have shown little interest in expanding. TSMC announced that it will cut its capital spending budget.
To fit these contradictory pieces of information together, it's important to remember that any stepper can print a whole range of feature sizes. It's possible to print even quarter-micron or larger designs using a 193 nm stepper. Costs aside, "wasting" process capability in this way will give improve CD control and yield, while simplifying the design problem.
The cost penalty isn't as severe as it might seem. For very low volumes, Sematech found [LINK], the cost of resolution-enhanced masks swamps the lithography capital cost. Larger feature sizes require fewer resolution enhancements, dramatically reducing mask costs.
Relaxed design rules reduce chip performance, but many designs can tolerate the trade of reduced performance for lower cost and faster time to market. Use of relaxed design rules doesn't necessarily contradict Moore's Law, either. Moore's Law says that circuit density doubles at constant price, but maintaining constant density at half the price also satisfies this condition.
Lithography has long been one of the most technically challenging segments of the semiconductor process. Below 100 nm, though, other processes face equally severe challenges. Even as lithographers make significant progress on both 157 nm and post-optical technologies, gate dielectric issues loom large for the 65-nm node.
Below 65 nm, researchers believe, leakage current in SiO2 gate dielectrics will become intolerable. Transistors must switch to a high dielectric constant (k) replacement material. Few candidates can match SiO2's chemical compatibility and ease of integration with silicon. Though metal gates can delay the need for high-k dielectrics, they introduce integration issues of their own.
A panel discussion on atomic layer deposition (ALD) presented the method as a possible solution to gate stack integration issues. As Tom Seidel of Genus explained, ALD is inherently a layer-by-layer growth process. It is especially well-suited to the molecular-level interface control required for very thin gate layers. Still, ASML's Jeff Kowalski reminded listeners that integration studies of potential high-k gate materials are only just beginning. Bringing an SiO2 replacement to production will, Dataquest principal analyst Dean Freeman said, "make copper and low-k look like a walk in the park."
Atomic layer deposition is relevant to process modules beyond the gate stack, too. Applied Materials' Endura iCuB/S, an integrated copper barrier/seed deposition system, is the most recent attempt to apply ALD to interconnects. The system uses ALD to deposit a conformal TaN barrier layer, followed by PVD of the copper seed layer.
Even before high-k gate dielectrics reach the production mainstream, thinner oxides are putting pressure on front-end-of-line (FEOL) cleans. Akrion's new aKros joined a parade of new clean systems aimed at the FEOL space. According to Akrion, the aKros's small process lot, small footprint, and low cost are well-suited to incremental capacity increases. At the same time, it allows superior process control with flexible recipes and chemical mix ratios.
Thickness, capacitance, composition, and leakage all contribute to gate stack performance. Thinner layers require more precise control of these properties, but are more difficult to measure. While optical thickness measurements are well-established, electrical measurements become more and more important after the 130 nm node.
KLA-Tencor's strategy for gate process monitoring combines both electrical and optical measurements. The Quantox XP measures capacitance and leakage independently. Though it is a non-contact measurement, its measurement spot size is too large for use on product wafers. Meanwhile, the Spectra FX 100 ellipsometry system can monitor thickness and reflectivity (which equates to composition) on product wafers.
Business consolidation emerged as a second major theme during the show. As companies become more confident in the emerging recovery, those with available reserves are aggressively pursuing growth through acquisition. In the week of Semicon West alone, Veeco Instruments and FEI announced a merger agreement, Rudolph agreed to buy ISOA, a macro defect detection specialist, and ATMI agreed to buy Microbar's copper delivery product line. MKS Instruments, Advanced Energy, and others have been steadily buying smaller companies for several months.
Many of the consolidating companies are small suppliers that make only one or a few products such as valves, power generators, or sensors. These companies are typically very good at what they do, have a small number of customers, and typically have less than a hundred or even less than 50 employees. With customers in Dresden, Taiwan, Oregon, and throughout the world, such a small company is likely to be spread pretty thin. Merging or being acquired is one way to address the problem. Another is to retreat to smaller, more manageable niches, like MEMs or compound semiconductors. As Trikon has found, other niches may also help cushion a company against the cyclicality of the IC industry mainstream.
Yet smaller companies are also a tremendous source of innovation. Applied Materials may be interested in new and better valves, but components are not the core of the company's business. Valves won't give Applied Materials a large enough return on its investment dollars. As small companies consolidate to meet the challenges of a global industry, a valuable source of component-focused "better mousetrap" innovation could be lost.
As one observer noted, innovation and new company formation tend to be cyclical. Engineers with new ideas become frustrated by the constraints of large corporations. They either strike out on their own voluntarily, or use post-layoff severance packages as seed money for new ventures. If a new venture succeeds, it eventually reaches the point where it is constrained by its lack of resources. It seeks additional resources either through a public offering or by being acquired, setting up another repetition of the cycle.
In general, people seemed more optimistic at the end of the show than at the beginning, which suggests that they talked to each other during the week and found that things aren't as bad as they thought. Equipment bookings and the equipment book-to-bill ratio have been rising for several months. Wafer shipments for the second half of 2002 jumped sharply relative to the first half. Assuming no further macroeconomic shocks, the industry appears to have turned the corner to recovery. The show week had the opposite effect last year, with people realizing that the downturn was worse than they feared.
| This site is Copyright ©2001-2002 by Thin Film Manufacturing. All Rights Reserved |