|
|
|
Low-K Race To Next Generation Copper rates highly as search on to supply semi needs San Francisco -- As some wafer suppliers skipped Semicon/West, materials attention turned toward copper interconnects, high-k dielectrics for capacity layers in memory circuits and low-k dielectric materials for buried layers. In the area of low-k dielectrics, chemical giants like AlliedSignal, Du Pont, Dow Corning and Amoco, among others, have been lining up to supply what is perceived as a next-generation semiconductor need. According to Sematech, signal delays are increasing as feature sizes are shrunk. Low-k materials reduce these delays incurred in chip interconnects. "Depending on designs and the complexity of your circuits, where these interconnection delays affect your circuits are at quarter-micron or less," says Ken Monnig, Sematech's associate director of interconnect. "You can reduce the line resistance, which is copper, or reduce the dielectric constant between the lines, which is low k." At Semicon/West, a top executive for a low-k dielectric material supplier privately said he was kicking himself for not promoting low k as a viable way of improving existing aluminum fab lines prior to IBM's copper process disclosure. Subsequent to IBM's 1997 announcement, copper has been the industry buzzword, but low-k materials, in this executive's viewpoint, could provide immediate benefits to existing lines. Not all executives and analysts covering the low-k materials field agree. Suppliers agree on the promise of low-k materials, while adding that copper has had more process discovery work. "I don't think low-k dielectric has matured as much as copper metalization," says Craig Schuckert, product manager at HD MicroSystems. HD MicroSystems is the Wilmington, Del.-based joint venture of Hitachi Chemical and Du Pont Electronics. "It (copper) is one material that has a lot of knowns," says Mr. Schuckert. By contrast, he added that multiple dielectric materials are being worked on today. "In reality, we thought people would start with low k and move to copper, but copper has turned out to be much easier to integrate and so people are going to move to copper and then integrate low k," says Mr. Monnig. Sematech started looking into low-k materials at the request of its member companies in 1993. The consortium took the unenviable task of "winnowing and screening" through a list of 155 materials and has collected loads of data on their performances and properties through rigorous measures. As the Semiconductor Industry Association roadmap guidelines draw closer, the degree of qualification and testing -- Sematech is trying to build two-layer metal, electrically testable structures with the proposed low-k materials -- become more extensive. Based on Sematech's data, members have narrowed the field of materials to roughly 10. Four or five relate to copper damascene technology, while another four or five are replacements in the subtractive metal etch process. Mr. Monnig was unable to give any specifics related to these finalists. He did say that Sematech is probably not going to look at anything above a 2.5 k constant, "although we will be building test structures above 2.5." Sematech is building these structures now and "will continue to do that probably through 1999, until we get as qualified a process as we have now with 2.5." Low-k suppliers agree that the market for k constants higher than 2.5 is not there. Chip makers are not as interested in materials with a 2.6-or-higher k constant and only somewhat interested in values between 2.2-2.5. While Texas Instruments and IBM have spent several years and millions of dollars on low-k research, most chip makers are waiting for further developments in the field, according to analysts. Most look forward to k constants of 2.1 and lower, materials such as porous silica and fluorinated organic polymer materials. "If a fluoropolymer is going to work," says Michael Mocella, a senior technical consultant at Du Pont, "Du Pont knows more about fluoropolymers than anybody else." Du Pont has several low-k dielectric materials that fall in the 1.8-2.1 range, most recognizable of which is Teflon AF. The brand-name material traditionally associated with no-stick cookware also can be applied to semiconductor wafers through spin-on technology. Teflon AF delivers a dielectric constant of 2.0. What remains to be seen is whether organic polymers can overcome the reluctance of chip makers. "So far, no polymer has been part of the IC structure," says Mr. Mocella. "Up to now, they have been entirely inorganic." "Not completely true," says Dan Rose of Rose Associates, an electronic materials market research and consulting firm. Chip makers have been using polyimides for years on the top layer of the chip. "Polyimides have never been used as a buried layer and that is sort of a mental barrier for the industry to overcome. It has been proven, but it is the reliability question that it has to overcome," he says. Dr. Rose adds polyimide reliability has been proven to a certain extent outside the fab; however, chip makers "have been using silicon dioxide forever, and are a little more comfortable with it." Since its base is silicon dioxide, something familiar to chip makers, porous silica comes closer to this manufacturer comfort zone. The material is formed by doping silicon dioxide to create bubbles within the material. This lowers the k constant. AlliedSignal has several low-k dielectric materials, including organic polymers; among these, the company has a porous silica material that fits the low-k (2.0 or lower) bill. Generically porous silica is referred to as xerogel or aerogel; AlliedSignal uses the name "Nanoglass." Its Nanoglass product measures 10-20 nanometers in diameter and has a dielectric constant of near 2. In March, AlliedSignal Electronic Materials consolidated its porous silica holdings by acquiring Nanoglass LLC, formerly a joint venture between AlliedSignal Advanced Microelectronic Materials (AMM) and Nanopore, Inc. (EN, March 23). Nanoglass LLC, which now operates as a unit within AlliedSignal, had been formed in 1996 to commercialize research done by Texas Instruments into porous silica for a low-k dielectric process. TI had been awarded a government contract to develop porous silica, and Nanoglass teamed up with TI to further the technology. Like organic polymers, porous silica has its own questions to overcome. "When you start to get down to about 2, you get to where the porosity gets to be a strength issue," says Mr. Rose. "There is a limit to how much porosity, how many bubbles, can be put in the porous silica before the structure will collapse. There is also a question of other things becoming entrapped in the bubbles," i.e., contamination. According to Mr. Monnig of Sematech, the toolsets can be completely different as well. Some materials require a spin-on process, while others require chemical vapor deposition. Among tool suppliers, Applied Materials and Novellus have low-k development projects. In May, Applied began the second phase of a program with Sematech to develop processes that can be used to etch emerging low-dielectric-constant materials. Sematech said it will use Applied etch systems to enhance knowledge of low-k materials and identify specific process issues related to etching these films (EN, May 18). Perhaps the most challenging element in low-k dielectric work is time. "We put 35 years into silicon dioxide and aluminum, and we don't have that level of maturity with copper and low k," says Mr. Monnig. Nor do materials and equipment suppliers have the time, given the SIA roadmap's aggressive goals for materials developments. If Semicon/West cemented one thing, it is that the next 10 years or so may prove to be a time of tremendous upheaval and excitement in this area. "We are going to make several materials changes in the integrated circuit structure each generation over the next several. That is something that we have never done," says Mr. Monnig. |