Christopher Case, CTO, BOC Edwards

Christopher Case Source: BOC Edwards

Christopher Case is the chief technology officer for BOC Edwards (Wilmington, Mass.), a gas and equipment supplier for the semiconductor manufacturing industry. He joined the company after spending 10 years at Bell Labs; during this time, he chaired the Sematech Interconnect Technical Advisory Board. Currently, he is the general chair of the IEEE International Interconnect Technology Conference, and for six years has chaired the ITRS Interconnect Technology Working Group. Prior to this, he held positions at Solid State Solutions, and was an assistant professor of engineering at Brown University. He holds a Masters from Université de Bordeaux , where he was a Fulbright-Hays scholar, and an Sc.B. and Ph.D. in engineering from Brown University.

SI: What trends do you see from your customers?

Case: For one, the industry is constantly asking us to reduce their costs as they try to meet productivity and customer pricing expectations. One way we can achieve this is by having improved products that either reduce CoO, or simply are less costly to acquire. No supplier has escaped this pressure.

SI: Other than the cost factor, are there any other similarities in what your customers ask for?

Case: Yes. Now that the implementation time for global warming guidelines is nearing, customers everywhere are starting to broadly recognize and address environmental issues. This is not to mean that they weren't aware of these issues in the past, but historically, many customers did only what was necessary from a regulatory standpoint. Often, that meant the minimum, which differed in regions around the world. However, now all our customers are extremely aware of the impending issues on global warming gases and other general environmental issues. Because we supply abatement equipment, and because our product platforms, as is the case with many suppliers, have an impact on the environment, we've seen a renewed interest in environmental factors. We're focusing a considerable part of our product development in the area of gases to address some of the newer process technologies.

SI: So what does your roadmap look like now?

Case: We're moving in concert with the times. For instance, most of our products are enabled with Ethernet connectivity, which allows remote control and monitoring of the system and can enable enhanced utility management. So under the control of the process tool, our equipment can go into a mode that reduces nitrogen or power consumption when the tool does not need its full capability — an idle mode, so to speak. We co-chair the SEMI E54 task force, working to specify an industry standard for interface between vacuum pumps and process tools. These are features we have been developing based on the roadmap's expectations that one must reduce, by a certain percentage, the tool's footprint and its power and water consumption each year. Few realize that as much as 30% of the power consumption in a fab can be accounted for by things like vacuum pumps, which, in turn, require considerable quantities of nitrogen to keep them operating from a cooling or dilution standpoint.

SI: So systems that do recycling must be growing in importance.

Case: Certainly. We have systems that can recycle gases: xenon, for example. This rare gas, less than 0.1 ppm in air by volume, is a byproduct of air liquefaction and is only produced in small quantities. In fact, the worldwide availability is fixed by the number of air separation units that capture it. Besides being used in some lamps, medical imaging and plasma displays, it has also been exploited as a good way to etch silicon wafers.

SI: But there isn't that much available.

Case: True. If just one of our customers switched their etching processes to xenon, the fab could potentially use 20% of the world's production, more when viewed from the perspective of excess capacity. Surprisingly, those experimenting with xenon for this purpose were relatively unaware of its scarcity, just as many are unaware of helium availability constraints. As people move from 200 to 300 mm fabs, typically the scaling of the gas usage is based on a combination of the volume of the chamber and the wafer's area. However, in some cases, there is overscaling to achieve process enhancements. In some of the LPCVD processes, we've seen helium and hydrogen consumption rise by a factor of 10 instead of the expected two or three. This has had an impact on helium's pricing, which also is a limited resource. Unlike xenon, which, except when it is used in rocket thrusters, eventually gets recycled back into the atmosphere, helium is a non-renewable resource. Once it is consumed, it is gone. So there is customer education involved, as well as reclamation techniques, to recover xenon or helium and recycle it back to the process tool.

SI: How do you view the ITRS?

Case: (Smiling) All roadmaps are wrong, but some roadmaps are useful. I have been involved with the roadmap since 1990, when the ITRS was called the NTRS — the National Technology Roadmap — and was based in the United States. Its focus was to help the U.S. industry regain its competitive edge with Japan. The fear then was that U.S. suppliers and manufacturers were losing to the Japanese. This is the same reason Sematech was created in 1988; it was a U.S. consortium for its first 10 years. Sematech and the NTRS helped drive the suppliers to improve the equipment they made by setting goals. Simultaneously, they wanted to set requirements from a technology standpoint to ensure Moore's Law was maintained. Sematech and the roadmap were very effective in the 1990s, when they actually achieved (some might say "overachieved") this result. One could argue that, in some cases, Sematech was too effective and created a set of large suppliers that now limits choice. We have some massive suppliers that, in many cases, provide a large percentage of tools in certain areas to both silicon and flat panel fabs. The fabs have lost some leverage because of this, and sometimes newer technologies take longer to get adopted.

SI: Do you still see the roadmap as useful?

Case: Indeed! It has grown in scope and is now in excess of 600 pages, plus hyperlinked material. It is probably the single, most heavily referenced industry document. The International Interconnect Technology Conference (IITC) draws about 800 specialists annually. I would have to say that nearly all of the papers presented reference the roadmap. It is key to keeping institutions focused on appropriate research, helps to ensure that it is something realistic that the industry might potentially adopt. There is some criticism that it is not aggressive enough, but, let's take an example: from a metal one half-pitch design rule standpoint, the roadmap forecast has tracked, with almost 100% correlation, what the leading-edge manufacturers have actually delivered over the last 10 years or so. In other areas, such as low-k dielectrics, it has not done such a good job. Here, the ITRS laid out a roadmap that the industry was unable to deliver.

The roadmap is being enhanced in the interface between the chip and the packaging world. Its job is to specify technology requirements, which are primarily needs-based. But, realistically, the potential solutions must help the industry stay on the traditional manufacturing path that delivers the expected productivity enhancements and cost reductions. We have also broadened the scope to include emerging materials and new devices.

SI: Scaling requirements have also changed.

Case: True. The industry has recognized that scaling doesn't have to be driven the traditional way of always making something 0.71× smaller every generation; that you can use equivalent scaling to achieve the same result. This is why some may choose to add an extra layer of metallization to achieve the same chip functionality, while others choose to save some money and shrink the transistor a little bit further. Companies can now deliver chips with more functionality, without necessarily having the same technology capabilities. Not everyone has to have the latest gate material, the latest transistor material, the latest dielectric or copper wiring materials. This is healthy.

SI: How would you summarize your R&D philosophy?

Case: This is something I am constantly assessing. I spent 10 years at Bell Labs and, certainly, for most of its history, it was world-renowned for its research. One could walk down the hallway with a question, and find someone who might be the world expert on that subject matter or a Nobel Prize winner. Anywhere from heat transfer to lasers, semiconductors, communications — anything! It was an extraordinary environment, but unfortunately it was not a sustainable one. It certainly wouldn't be so in today's changing, competitive world. In the '80s, chipmakers continued to invest heavily in R&D. In the '90s, some of that role was offset to the largest OEMs, which were able to offer complete process solutions. Today, we are seeing a renewed but targeted investment by the major chip manufacturers, but they are also embracing a new model called "open innovation" in which the challenge of doing research is shared with several participants. The people who did the fundamental research have all but disappeared from the companies that used to have them. We had 3× as many Ph.D.s doing research in just one New Jersey location than we do today. The prior level of investment was no longer sustainable. Now, we have more programs with universities, work closely with consortia, have more joint ventures and activities with customers and even competitors, and participate in many government grants.

SI: Research has become too challenging and expensive?

Case: Yes. And a good example, again, is low-k materials. When the roadmap first mentioned the need for low-k materials, it was 1991. In the next 10 years, companies spent over $100M developing these low-k materials. Of those companies, if they still exist, only a few remain committed to low-k. Some might say that the roadmap was driving the industry the wrong way, but it was really the investment required not only to develop the materials, but to perform the tests and integration, chip evaluations, and everything necessary to prove the concept — all too costly for the materials OEMs. The technology has continued to challenge us all, and now even the large OEMs no longer have the skill set to do the integration needed by manufacturers. There's a renaissance taking place in the mainstream chip companies — leading to a renewed interest in R&D. These companies are bringing back in-house some of the R&D they had previously passed on to the OEMs. But they are being more selective and more effective.

SI: Just another cycle?

Case: Yes. Look at the Japanese, for example. They were the leaders in DRAM in the 1980s, then South Korea put competitive pressure on them, and they did not make the transition and experienced an economic slump. They missed the train on 300 mm, and were late on copper, but they are now presenting some of the best technology papers at the leading conferences. As a country, they've also focused on putting consortia in place to bring together Japanese manufacturers and OEMs to increase R&D leverage. I believe we will see two Japanese manufacturers deliver production-qualified, extreme low-k logic chips in 65 nm technology before anyone else. I think we will also see Korea and other Asian countries develop some very capable global OEM suppliers.

SI: What do you see as the main challenge over the next five years?

Case: In this year's roadmap, we will acknowledge that the microprocessor and logic roadmap will likely accelerate to a two-year cycle — about a 0.75 shrink every two years for the next two generations. As an industry, we will deliver a true 90 nm logic product this year, and for the next two cycles, until 2009, we'll remain in a two-year cycle. This is an acceleration from what, over the last several nodes, had been a two-and-a-half or three-year cycle. Then, around 2009, both DRAM and logic will relax again to a 0.7 shrink every three years. From a local or first-level metallization standpoint, this will bring a coincidental convergence of this same key technology driver, and we will follow a unified first-level metal pitch design rule.

SI: What effect will this have on the industry?

Case: Already, there is a lot of memory embedded on logic products — SRAM — and ultimately the goal of chip designers is additional functionality. This was behind the concept of SoC (system-on-a-chip), which in many cases transitioned to system-in-a-package. If we can merge the functionality of DRAM, SRAM, logic and generic ASIC, along with the extra functionality for things like optical emitters and receivers, RF devices and MEMS on die, we will give manufacturers what they've desperately needed for years: a way to sell their chips for more than the cost of the silicon. The way a chip is quoted by a foundry is to determine which technology and die size is required, then a price is quoted based on die size, not function. So, except for a few companies with a strong share in microprocessors or other high-performance chips, for example, the chip price is not much more than the manufacturing cost of the die.

SI: Ours must be the only industry where this happens.

Case: It is. A square centimeter of die with great capabilities ought to be worth more than $5 to make it. We must give the chip industry a way to price chips based on functionality, not area.

SI: Not everyone would want that — the cost of consumer electronics would skyrocket.

Case: True, they might rise. I bought another DVD player at Frye's Electronics recently. It is multivoltage, multiregion, progressive scan, component output, compact footprint — $19. And it wasn't even a special sale! An extended warranty would cost more than the player itself. I don't see how one can even pay for the packaging and inventory, let alone for the cost of the components. It's wrong. Consumers are being misled into expecting products to be this cheap, and we've developed the philosophy of, if it doesn't work, toss it and get a new one. There must be a better way. We're the highest-tech technology industry ever. Every other high-tech industry I can think of uses ICs, yet we do not have much margin left in our products. This starts with the supplier of sand all the way to the manufacturers making those chips.