The future of ULK dielectrics
04/01/2005
Solid State Technology asked industry experts to discuss ramifications and issues concerning ULK dielectrics development.
A case study: Highly porous ULK dielectrics
The solid consensus across a panel of R&D researchers at last October’s Sematech-sponsored Low-k Workshop was that ultralow-k (ULK), highly porous intermetal-dielectric (IMD) films would not be used until about 2009. Accounting for the expected optimism of researchers, and focusing on high-volume production, it seems reasonable to expect that ULK will not be ready until 2013 or so. This date presumably bears relevance to significant revenues for materials suppliers and equipment OEMs. There was broad agreement that these second-generation low-k materials would find slow implementation, and at a technology node that itself would be slow to develop [i.e., 35nm in International Technology Roadmap for Semiconductors (ITRS) terms].
This should pose a serious dilemma for the extraordinarily large number of OEMs and materials companies that are either engaged or are considering engagement in ULK product development. A dollar spent in product development - starting in 2002, for example - cannot logically lead to profitable income for 11 years or so; and technical, market, and competitive risk factors are huge. The 1993-2000 investments have long since been written off.
Clearly, the suppliers making these investments aren’t ignorant of the facts, and the true experts within the IDMs have consistently expressed great pessimism (although they certainly encouraged the investments). So what has driven the nine-figure dollar amount that has been cumulatively invested thus far? In 1993, while managing the first serious highly porous ULK materials project, I began dialogues with IDM experts and a quest to understand the common, detailed technical concerns of these interconnect leaders. Virtually none of these experts believed that glass or plastic highly porous “sponges” would replace hard, inert silicon dioxide as IMDs. In my view, unrealistic optimism has come from suppliers with a vested interest in highly porous ULK R&D. Expensive porous ULK projects over the past decade often have been funded with minimal scrutiny for technical and market risk.
Achievement of lower keff values remains stalled by multiple technical hurdles associated with bulk low-k films, including improved etch stops and cap layers, and process integration into high-volume manufacturing lines. The drive to achieve cost efficiency based on standardization dictates that, at most, two or three ULK solutions will eventually be adopted; yet it seems that for every materials company that drops out, two more enter the fray. On the development side, for every porous material deficiency that is overcome, two more are seemingly identified - including new failure mechanisms found by Sematech and others.
The enterprise of using highly porous ULK dielectrics is akin to building the Space Shuttle from balsa wood: great from a payload standpoint, but it is simply the wrong material. Industry experts recently have made the following point: By the time you implement the supporting films and strategies to actually use a highly porous IMD, you invariably are left with a much higher keff than you were targeting, to say nothing of the increased film interfaces, complexity, and cost.
On the other hand, if you process (etch/ash/clean) a highly porous dielectric film in a more or less conventional way, the damage done to the film raises the k value such that it renders the film unusable. If a porous ULK IMD ever makes it into high-volume production, I’ll be surprised. If it happens in less than 8-10 years, I’ll be shocked. One hopes that some novel and potentially satisfactory alternatives to porous dielectrics are in development. Recent technical conferences provide some reason to believe this. There are interconnect capacitance-control strategies that don’t rely on ULK “glass sponges,” but none that as yet offer low-cost, high-volume manufacturing.
So what is the status of advanced IMD development? First, the slowing of Moore’s Law will provide more time than usual to find a replacement for first-generation keff = 3.2 interconnects. Second, either a modified air bridge or a novel solid ULK material will gradually be implemented. Finally, a significant emphasis on systems performance, as manifested in chip-to-package signal improvements and 3D (stacked) interconnects will, to some extent, mitigate the need for reduced keff in chip interconnects.
 |
For more information, contact Neil Hendricks at Advanced Wafer Fab Materials, Santa Cruz, California; e-mail [email protected].
ITRS: In a holding pattern with ULK
The ITRS is a remarkable document, practically unique among all other industry sectors, and a real credit to the hard work and critical thinking of the hundreds of industry experts who contribute to its publication. In the past, suppliers, universities, and governmental agencies have successfully focused resources to solve many intractable technical problems because of this document, but the low-k dielectric challenge was, and still is, an entirely different order-of-magnitude in complexity, compared to many previous issues. The low-k dielectric challenge has been different in at least three important ways.
First, low-k dielectric materials are an overall integration problem, not merely a simple one-for-one materials substitution issue. Consider the most sophisticated materials used in IC fabrication - DUV photoresists, STI slurries, etc. - then think about the number of process steps that require compatibility with these materials. One, or two at most, is the answer in most cases. In contrast, the dielectric - whether low-k or not - remains behind for the hundreds of subsequent processing and packaging steps as well as device reliability testing, and ultimately, deployment in the field with the finished chip. Everything has to work together perfectly to pull this off, so the focus on a simple k-value number on the Roadmap dangerously oversimplifies the challenge.
The second major difference in low-k implementation is not technical at all. Politics, momentum, and the culture of each individual customer tend to drive decisions as big as the next dielectric platform, and there are armies of influential vendors and consultants out there to help companies make the right choice. Dow’s SiLK semiconductor dielectric, for example, was proven to work technically by IBM and is in volume production today at Fujitsu, but got shut down at 90nm when industry momentum seemed to shift in a different direction. No roadmap could have predicted this, any more than it could have predicted the overnight shift the industry made last year away from 157nm lithography and toward immersion lithography.
Finally, in an industry as creative and competitive as ours, substitution technologies and approaches can unexpectedly spring onto the scene to achieve the same goals as those prescribed in the Roadmap. Low-k materials originally were postulated as the best means of decreasing IC price/performance, but alternate approaches in chip design and layout and technology breakthroughs such as strained silicon have subsequently achieved the same targets. The industry will stay with the current k~3.0 CVD solution for as long as it perceives that the risk of going lower in k value is higher than the alternatives available for achieving required next-generation chip performance, regardless of what the Roadmap says.
As good as the ITRS is at identifying, isolating, and focusing resources to work on specific “red bricks,” we expect that ultralow-k (k~2.2) will be an even more intractable challenge than low-k is today. This is again due to the daunting complexity of the overall process integration, the availability of alternate technologies to accomplish the same price/performance point, as well as the political risk of implementing a solution that may not be perceived as mainstream.
Our expectations of the ULK market are that:
- No totally new “breakthrough” ULK materials will be discovered that are not already known or under consideration today. This is because ULK - like low-k - is not solely a materials problem, and the cost to solve the integration, political, and technical risk aspects of the challenge are simply too high for any new supplier to bear.
- Until a technical leader emerges from the IDM or foundry community and demonstrates that low-k can also be low-risk, the industry will be stuck in a holding pattern - despite what the latest ITRS calls for - with the current k~3.0 CVD materials, and a few with SiLK resin at k = 2.6.
The ITRS has been and, to a large extent, still is a useful document for most issues. As lithographic scaling gives way to increasingly complex challenges, however, the ITRS will have to evolve with it to retain its impact in the coming “decade of materials.”
Acknowledgment
SiLK is a trademark of The Dow Chemical Co.
 |
For more information, contact Mark McClear, global business director, at The Dow Chemical Co., 1714 Building, Midland, MI 48674; e-mail [email protected].