The Worthwhile Pursuit of Low-k Dielectrics

The semiconductor industry's challenging pursuit to develop high-performance low-k dielectric technology at the 130 nm node has been somewhat like Thomas Edison's charged mission to invent the lightbulb. It took the famous inventor more than 1000 tries before he got a bulb to light.
 
I don't mean to imply that developing low-k film material that works with 130 nm technology is as big an invention as the lightbulb. But it's an important milestone in the ongoing evolution of semiconductor technology performance improvement. Like Edison, along the way the industry has faced tests, starts and stops, and missteps before it has gotten the technology to work right.

This year, after this protracted struggle, the mission was accomplished. Approximately a half dozen companies offer 130 nm chips with low-k dielectrics. Equipped with this technology, faster and lower-power chips can be produced than those currently on the market that don't use low-k dielectric technology.

In fact, earlier this year, Taiwan Semiconductor Manufacturing Co. (TSMC) production-qualified the first communications chip using low-k dielectric 130 nm technology. The chip is being used in third-generation wireless base station equipment.

Developing this low-k technology has revealed the rigorous expertise required to make it commercially available. Necessary expertise includes end-to-end system design, packaging, stress testing, assembly, reliability, understanding of the stability of low-k dielectric films, and chip manufacturing process technology. Manufacturing the chips has not proven to be necessary to lead in this major technology race.

Working with a manufacturing foundry, a "fab lite" business model has produced leading-edge technology ahead of companies that manufacture their own chips. Such a business model, in this context, has delivered the benefits of both worlds: leading-edge technology with a foundry, and economic benefits of dramatically reduced fixed manufacturing costs.

Four to six years ago, when this technology landed on the industry's radar screen, the conventional wisdom held that low-k dielectrics would probably work without too much difficulty. However, the industry continues to wrestle with some challenges even now. And the technology is not as broadly used as originally anticipated.

One mechanical challenge has been ensuring the low-k film has the right dielectric constant to enable electrical signals to pass at higher speeds through the transistors, without increasing mechanical flaws, such as film cracking. As the dielectric constant levels decrease, and the mechanical properties of the material change, the more likely it is that the dielectric material will crack. Such cracking reduces reliability.

Although most of the industry remains focused on rolling out 130 nm chips, it is likely low-k dielectric technology will be much more widely used for chips at the 90 nm technology node. As a result, semiconductor companies must keep pace with this trend by figuring out how to make it work at 130 nm so they can progress to the 90 nm node. If they don't keep up, they could find themselves out of business.

As with 130 nm low-k, the types of chips that will use 90 nm technology are expected to include traffic management network processors, switching chips, application-specific ICs (ASICs), framers, mappers, and clock synthesis chips.

There is no question that many chip companies would like to work with low-k technology. But it's important that they remain technology-agnostic. Companies should be mainly interested in any process technologies that enable them to deliver to customers the highest-performance chips at the lowest possible costs and highest reliability. Currently, this technology happens to utilize a low-k dielectric at 130 nm for some applications. For other chips, semiconductor companies might decide to use other process technologies.

Still, low-k dielectric is clearly a viable technology with strong potential for extensive use during the next several years. And the industry has taken several substantial steps to make that happen. Granted, the use of low-k dielectrics is not as groundbreaking as inventing the lightbulb. But it's important because the chip industry thrives on improving performance, which is what using low-k dielectrics is all about. The industry should feel pleased for plowing ahead as far as it has to date, and view the future of this technology in a positive light.