Roadmap Paves Many Paths in Front-End Processing
-- Semiconductor International, 2/1/2000
Architects of the 1999 SIA ITRS call the transition from today's silicon dioxide/dual-doped polysilicon gate stack to a high-k/dual-metal gate stack a potential " showstopper' that will require significant resources to overcome." The CMOS gate stack transition is expected to be useful for two device generations beginning in the year 2005. Interestingly, of equal importance is the transition to alternative high-k dielectrics with metal capacitor electrodes from today's nitride/polysilicon capacitors, a process posing "problems of equivalent complexity and scope as those associated with the future CMOS gate stack," with, unfortunately, a similar timeline for manufacturing deployment. Along with these monumental changes are the requirements for abruptly doped ultrashallow junctions with dopant concentrations in excess of their equilibrium limits; tighter etch CD control; reduced thermal budget; and silicon substrates (including 300 mm) of purity necessary to maintain high device yield.
Tunneling current precludes the use of SiO2 dielectric layers below approximately 1.5 nm, pointing to a phase-out of oxide gate dielectrics as early as the 100 nm node. Such solutions as oxynitride and Si3N4, which impose restraints on surface preparation, ambient control, thermal budgets and silicon compatible materials development (for contacts and gate electrodes), are capable of extending the technology about five years. Beyond the nitrides, medium-k (10-20) gate dielectrics of Ta2O5, TiO2 or ZrO2 are being considered, as well as silicates of Zr, Hf, La and Ti, followed by high-k (k>20) amorphous materials of LaAlO3 or ZrTiO4 or single-crystal LaAl3O4, BaZrO3 or Y2O3. All these materials are in the research stages of development. The Roadmap emphasizes that the full benefits of the higher-k gate materials can only be realized if the ultrathin interfacial layer of SiO2 can be eliminated, which currently appears necessary. Beyond the dielectric, traditional poly gate technology may be extended through enhanced dopant activation. One way of achieving this is by doping the silicon with germanium. No later than 2005, dual-metal gates must be integrated with the higher-k dielectrics. Metal gates of Ta, Zr, Hf and Ti are being considered for NMOS transistors with TaN, WN, Pt or Ir for PMOS transistors to achieve acceptable threshold voltages for each.
Conventional salicide processes are challenged by the conflicting demands of producing highly doped more shallow source/drain (S/D) and drain extensions and contacting them with low-resistance materials. Activation technology is needed to achieve 300 V/sq resistivity at <30 nm junction depths. Other issues in silicide processing include prevention of boron penetration through gate oxide, keeping particle levels low compared to junction dimensions and high doping precision, particularly with halo implants. The device is becoming increasingly sensitive to damage from plasma etching and ion implantation processes after gate formation. If and when elevated S/Ds are implemented, new sidewall spacer approaches must be found. However, elevated contact structures are only in the research stage.
Along with the scaling of S/D junction depths comes the need for abrupt retrograde wells. Such vertical channel engineering addresses increases in off-state leakage that accompany the scaling of threshold voltage. As channel lengths scale to the sub-50 nm range, lateral channel engineering using angled implants to create dopant variation across the channel also becomes more crucial.
Metal gates and deposited high-k dielectrics drive etch CD requirements for low bias and high uniformity of edge profile across increasingly larger wafer diameters. Thinning photoresist masks are leading to the use of hard masks for pattern transfer, films that might also serve as antireflective coatings. Selectivity to underlying films becomes a tremendous challenge with new materials. Charge damage also is a concern during plasma etching (with chemical downstream etching as one option), as is defectivity. "To meet future defect density requirements the plasma processes and etch tools must generate considerably fewer and smaller particles." Measurement techniques to determine the remaining oxide thickness after gate etch cleaning are becoming critical.
Based on a cell capacitance of 25fF, stacked DRAM capacitors are expected to require Ta2O5 dielectric that, when combined with polysilicon bottom electrode (metal insulator silicon (MIS) structure) at the 180 nm node realizes a k=22 due to oxide at the poly interface. Full advantage of Ta2O5's k=50 dielectric constant results from its combination with alternative electrodes of Ru or Pt in a metal insulator metal (MIM) structure at the 130 nm node. Metal or metallic oxides such as Pt, Ru, RuO2 or IrO2 will make up the storage node bottom electrode at 130 nm, restricting the use of high-temperature annealing. To reduce the aspect ratio of the storage node and contact hole at the 100 nm node, BST dielectric (k=200-250) will be needed, along with new cell architectures to continually shrink cell size. Contact hole formation also poses problems for system-on-a-chip designs as DRAMs typically have much deeper contact holes than logic components. Trench DRAM capacitors will not require such dramatic material changes as the stacked designs, with extension of the NO dielectric to the 70 nm generation.
Finally, in the area of starting silicon, the Roadmap calls for a "methodology to control overall material-contributed yield loss by partitioning loss allocations by defect type so that an overall IC fabrication yield loss due to starting material does not exceed 1%." In this way, semiconductor manufacturers could determine acceptable product distributions and use them as materials acceptance criteria, enabling a fit of specifications to product distributions demonstrated by the wafer suppliers. Today's approach differs in that critical characteristics of given wafers are listed such that each parameter contributes no more than 1% to chip yield loss. The values are derived from model-based analysis, whose models are "incomplete," according to the Roadmap, "as they do not take into account the distribution of parameter values resulting from manufacturing variation." Other highlights in the silicon area include the increasing importance of SOI beyond a niche technology, need to continually reduce surface residues such as particles and organics, and increasing attention being paid to wafer backside contamination levels and flatness.
Despite the many material changes and associated integration challenges deemed necessary in the very near future, the Roadmap architects offer encouragement: "Although fundamental limits are being approached for the materials and devices currently used, these same limits may not apply to new materials and devices." •
SCALPEL Provides Best Strategic Fit for Lucent
 |
| David Eaglesham. (Source: Lucent Technologies) |
David Eaglesham is director of the Silicon Electronics Research Lab at Lucent Technologies' Bell Labs (Murray Hill, N.J.). He joined Bell Laboratories as a member of the technical staff in 1988. A graduate of the University of Bristol (England) with a Ph.D. in physics, he has authored more than 180 papers on implantation, diffusion and crystal growth mechanisms.
SI: With the industry beginning to turn around, do you see changes accompanying the recovery?
Eaglesham: Despite the slow turnaround, I think business is in good shape. The slow pick-up is partly a result of the $700 PC phenomenon. But we're seeing continued strong growth in the communications sector, and that's becoming increasingly important across the entire IC business. What's occurring is a change in what's driving the IC industry, from the PC toward the communications sector.
SI: How does that impact IC manufacturing?Eaglesham: Historically, the microprocessor was a primary revenue driver. Now that we're seeing a move toward communications, this calls for a slightly different emphasis both on the technology and on the business side. For example, in the microprocessor segment, transistor drive current is the most important driver for technology. If you're selling a 600 MHz chip for 3 times the price of a 550 MHz, then cost becomes a secondary part of the equation. However, in the broader IC industry, cost per function is the dominant driver. The technology required to support the two segments appears to be diverging. It's possible both the transistors and the manufacturing tools will be different for the two business segments.
SI: Is this change in any way related to SEMATECH's recent announcement to support multiple next-generation lithography (NGL) technologies?Eaglesham: I think there's a relationship there. In practical terms, SEMATECH has been trying to get a consensus to arrive at one NGL technology, but instead there's been a consensus to support two technologies. EUV technology is a performance-at-all-costs approach, and is most strongly supported by the segment of the business where cost is much less important. The segment that needs to drive cost per function, as well as performance, is driving SCALPEL-based lithography.
SI: How can the industry support multiple technologies?Eaglesham: I believe that's a question for the tool vendors, more than for the IC companies. The way it's playing out at the moment is that the two largest lithography companies, Nikon and ASML , and the largest tool supplier, Applied Materials, ar e all developing tools for electron projection lithography (EPL), SCALPEL-based technologies. On the other hand, Intel is paying for EUV development. If it's in Intel's best interest to drive the industry infrastructure for EUV, then they will continue to do that. The cost-performance segment will continue to buy the best tools available.
SI: Among the NGL technologies, why is Lucent supporting SCALPEL ?Eaglesham: It's primarily cost — specifically mask costs. The specific problems that most companies have with other NGL technologies are in high mask costs. And the impact that has on your business depends very critically on the nature of your business. If you're selling a one-size-fits-all component -- say, millions of microprocessors -- then mask costs are less significant than if you're in a fast-moving business like communications, where flexibility is everything. For most of the IC business, mask costs dominate the overall cost of lithography. With SCALPEL, mask costs could be an order of magnitude lower than the competing technologies.
SI: One concern with SCALPEL has been space charge effects, where beam blur essentially creates a trade-off between throughput and resolution.Eaglesham: That's true for a fixed numerical aperture (NA). If you keep the column design constant and the NA fixed, you need to make a trade-off. However, as with improvements in optical lithography, successive generations of tools should improve, and a small increase in NA will more than offset the decrease in throughput.
SI: Given the risks involved in all the post-optical lithographies, is Lucent supporting other alternative technologies?Eaglesham: The EUV program started at Bell Labs, and we've had very significant programs in X-ray projection, X-ray proximity and ion beam lithography. We've looked at the technical risks. And at this stage, we're backing the technology that has the best strategic fit for our business, and that is SCALPEL. But at whatever technology node, we'll buy the tool that does the job best.
SI: Looking ahead to the next five years, what do you envision?Eaglesham: I anticipate a very serious look at alternative device structures. There are some big challenges in continuing simple scaling. There have been articles from Intel and IBM announcing the end of the Roadmap, and they are not alone. At Bell Labs we've been much more optimistic and have announced ways to circumvent many of these challenges. At the IEDM meeting this past December, there were a number of papers where people are looking at novel transistors as a way around further geometry shrinks, and the Bell Labs device work was very well received. The consensus at the meeting was that people will be looking at alternative device structures as well as some changes in materials. I don't anticipate seeing major changes in manufacturing within five years, but programs will mature, and advanced device designs such as vertical transistors will come much closer to commercialization. • -- Ruth DeJule
EKC Licenses New Photoimaging Process
EKC Technology (Hayward, Calif.), obtained exclusive rights under a license with Simon Fraser University (Canada) to commercialize a patented process that uses light in combination with metallic compounds to directly deposit metal and metal oxides on a substrate. Advantages of the process include low-temperature (<200degC) processing, simplified process flows, elimination of vacuum processing steps and lower capital costs. Potential applications range from printed wire boards to advanced ICs. The process, set to be in beta by the end of this year, addresses many material and process challenges identified in the 1999 Roadmap. •
Akrion CEO Revisits Trek Out of Corporate Oblivio
David J. Ferran is CEO and chairman of Akrion Inc. (Allentown, Pa.) and previously was president, CEO and a director of SubMicron Systems Corp. Before that, he served as president, CEO and chairman of Tylan General Corporation and president of its predecessor, Vacuum General. Tylan General went public in 1995 and later was acquired by Millipore Corp. Ferran has a B.S. in Business Administration from the University of New Hampshire.
 |
| David J. Ferran, Akrion CEO and chairman. (Source: Akrion) |
SI: You have an interesting story to tell about the rise and fall and rise of SubMicron, and Akrion's birth from its ashes.
Ferran: [Laughing] Isn't that supposed to be a Chinese curse, "May you live during interesting times?" I certainly have had my share of those. My career in the semiconductor equipment side began when I became the co-founder of a company called Vacuum General. Despite a minimum of experience, we grew it and ultimately bought Tylan in 1989 and formed Tylan General, which we took public in 1995. We were very successful until we came became targets of a hostile take-over. Ultimately, the company had to be sold to protect it from the take-over, with Millipore as our white knight.
SI: Did you stay with the company after it was sold?Ferran: No, after 18 years of hard work I decided to take a year off.
SI: Did you take the whole year?Ferran: Tylan General was sold in February 1997, and I got a phone call from SubMicron's COO six weeks later. They were looking for a new CEO. The board decided a management change was required, and he thought I might be the one to turn things around. I joined them in early May.
SI: What prompted you to make the move?Ferran: After meeting with some of the management people, it became clear to me the problem's origin was the management philosophy, not the technology. I believed I could correct a management problem, whereas a technology matter wouldn't have been my strong suit. I realized it wouldn't be easy, since the company was close to bankruptcy.
SI: Bankruptcy? How critical was the situation?Ferran: When I joined SubMicron, they were in default of $28M in bank loans. The previous year they did $178M in revenues and had 675 employees. They were publicly traded, and their stock price had dropped from $14 to $4. I joined on May 12, and my first meeting was with the consortium of banks that SubMicron was in default with. They started the meeting by telling me I had until month's end to repay the loans or they were pulling the plug.
SI: Were you aware of the financial situation before coming on board?Ferran: I knew they had bank troubles but figured that if the company shut down, there'd be little cash left for the banks; so I thought I'd be able to convince them to give us more time to put together a plan to pay them back. However there were several other serious problems I knew nothing about.
SI: Were you successful with the banks?Ferran: Yes. I presented the plan, told them we needed time to refocus the business on its core competencies. We planned to sell off some business units that were not in our core area of wet surface preparation, and I was counting on those sales to generate cash to pay the banks off. I also said that if the sales proceeds were insufficient, I'd raise new money to pay them.
SI: Did it work?Ferran: Not exactly as I'd have liked. Between May and November 1997, we sold off two business units, but it wasn't enough to pay off the loans. I was able to find new investors to furnish us the balance of the money we needed for the banks, while providing us a modest amount of fresh working capital. But it was expensive money, and the interest expense hurt us.
SI: Did the investors come in because of you or the company?Ferran: [Smiling] They knew me (and benefited) from the Tylan General days, and were willing to provide the $20M in additional funding I needed. Believe me, it wasn't based on SubMicron's track record! I'm sure of this because we approached and were turned down by nearly 50 other financial institutions.
SI: What were the problems you were unaware of?Ferran: The biggest surprise to me was the complete lack of systems and disciplines that I took for granted at my previous companies. They were missing all the basics such as mission, value, quality and vision statements. There was a lack of financial planning and control, and product development and manufacturing operated without even a minimum level of systems and procedures. The more I got into it, the more I realized the fundamentals of running a semiconductor capital equipment business had never been implemented.
SI: No annual operating plan, budgets, vision or value statements?Ferran: Incredible, isn't it? I was dumbfounded that a publicly traded company their size wouldn't have any of these things. Digging deeper, I discovered they didn't even produce monthly income statements! They only generated the quarterly statements the SEC requires. Further, the financial information available within the company was not regularly disseminated to management below the top three executives. Accounting expertise was at a minimum, and the numbers didn't reflect what actually was going on. I didn't fully realize this until months after joining. They had receivables out over 700 days, mostly with customers who were completely dissatisfied with equipment quality and performance.
SI: Didn't you talk to industry people before, to see how they were perceived?Ferran: Indeed I did, and they all warned me to stay away, saying they thought management didn't deal very honestly and they would never buy from SubMicron again.
SI: No offense, but you sound like a frustrated kamikaze!Ferran: [Laughing] It certainly must look that way! It sounds corny, but what did it for me was the challenge, the opportunity to use 18 years of learning and being in management. I believed that if I could focus the SubMicron people on the right things we might find a way to save the company.
SI: What were your first steps?Ferran: I talked to customers. They all thought the technology was sound and innovative, but the product design and execution were deficient. Deliveries were always late, and when the wet cleaning station arrived it was in pieces, looking as if it were being designed on the floor. They complained installation took forever and equipment didn't meet specs. Reliability -- the MTBF -- was miserable; machines were difficult to service, and the service itself was described as "lousy."
SI: Hadn't SubMicron management tried to correct these problems?Ferran: Not at all! When called by unhappy customers, management was arrogant and unwilling to work with them. Mind you, I got this from everybody-- all the customers I contacted, which included National Semiconductor, VLSI, Motorola and others.
SI: Were there any favorable comments?Ferran: Yes. Several said that when the equipment worked, they got process results they couldn't get on any other wet bench. But I realized the product wasn't engineered to be manufacturable, easily installed, reliable or easily serviceable -- and it sure as hell wasn't engineered to be profitable! Despite this, the core technology, the core chemistry, the fundamentals of that chemistry and the megasonics and the tank design, the fluid dynamics, those were solid. When the beast worked, the results were often terrific.
SI: What about inside the company?Ferran: At the top of SubMicron there simply was no interest in satisfying the customer; it never came up in my discussions with senior management. All I heard about was the stock price, press releases, revenue growth, acquisitions and the value of their stock options. I never heard anything about the customer, employees or suppliers. The bond holders were particularly irate and felt they had been misled about the company's financial situation and prospects before investing their $18M. Within months of making the investment they knew they'd been had. Ultimately, they lost everything. Many still believe someone should have gone to jail.
SI: What did you do when you took over?Ferran: On my first day, after meeting with the banks, I asked the head administrative assistant to set up a meeting of the executive committee. She didn't know what I was talking about. Just then the COO walked by, and she asked him what I meant, and he asked me in turn. I said I wanted to meet with the functional managers, the people who ran the business. He said they did "not have an executive committee or whatever." I asked who ran things, and he replied that the former CEO did. I said one person couldn't do it all alone (I sure as hell couldn't!), and you needed a management team to work out issues and run the business. "We don't do that." I then asked how people knew what to do, and was told everybody did what they thought they had to do, and if it was wrong the former CEO would "set them straight on the plan." I said I hadn't seen the plan and asked what it was, and was told it was so simple it didn' double sales every year.
SI: How can a company run that way get to be publicly traded?Ferran: The founders had some very good ideas about automated wet cleaning systems. Before SubMicron, wet benches were just little more than plastic dishwashers-- low-tech, with little process control and automation. SubMicron brought a technology focus to wet benches. During 1991-93 the industry was strong; they caught the wave and got some large orders. They never managed to deliver on their promises or satisfy customers, but they had the orders and they grew. Since they didn't have in place any of the systems or disciplines necessary to go public through normal means, they arranged a reverse merger into a public shell and went public without the public scrutiny.
SI: How could they grow, acting as they did?Ferran: They used proceeds from the public offering and their newly tradable stock to buy companies and technologies. Most of the growth was acquisition-fueled. Despite significant top-line gains, the company never generated much profitability; but there was the allure of profits to come, and management kept pointing to past growth and projecting future growth, and promising profitability-- it was all smoke and mirrors.
SI: What happened next?Ferran: I called the first all-hands meeting in the company's history and had a heart-to-heart chat with all the Allentown-based employees. I told them I wasn't sure SubMicron could be saved, and that to do it we had to do things differently. They posed questions and got honest answers. I told them to honestly self-assess their performance and asked, "When it comes down to taking care of our customers, shareholders, suppliers, employees and the community we do business in, how many of you believe we have been an A' performer, B,' C'?" No hands went up. I joked about quitting, and everyone laughed, but now I was scared to death.
SI: How did you begin correcting this?Ferran: I took 22 people, one from each of the company's functional areas, to a three-day offsite retreat. We began by putting together a value statement. I asked them what words they'd use to describe the company if they were outsiders. "Sneaky, dishonest, arrogant" came up. Everyone felt that in working for SubMicron they had been forced or encouraged to compromise their integrity. We finally produced a statement, put it in the lobby, signed it and told the rest of the employees to sign up when they thought the new management was really committed to meeting these values.
SI: What was the statement?Ferran: "SubMicron employees are committed to excellence and innovation in everything we do. We operate with integrity and professionalism in our relationships with our customers, suppliers, shareholders, the community and each other." That began the change of the company culture.
SI: When did you perceive things were beginning to work?Ferran: When rank-and-file workers started coming directly to me to share concerns, complaints, ask questions and offer suggestions. One day, our director of materials came to tell me he'd crossed his name off from the value statement because of some decisions he couldn't support without compromising his integrity. We called a meeting of everyone concerned, and it turned out to be a misunderstanding. We spent quite a long time correcting the matter, and after the meeting he re-signed. People were taking the values seriously.
SI: What about your customers?Ferran: I personally met with those who were upset about our tool performance and told them I knew they felt lied to and cheated. I asked them to allow us to upgrade their tools and do everything necessary, at our cost, to bring them to a higher performance level. A number of these companies are now Akrion's top customers. Others said they didn't want us back under any conditions, but they're now slowly coming around.
SI: Those first meetings must have been rough.Ferran: Quite an understatement! I never came across the personal animosity that was prevalent between SubMicron and its customers. We explained the new management's commitment to correct past mistakes, and embarked upon a reliability upgrade program to improve the tool and reduce installation time. We built up a stronger field presence and parts support capability. We hoped to hang on until an industry upturn came and get orders, but early in 1999 it became obvious the industry wasn't going to rebound in time. Our lenders and investors weren't in a position to provide any additional financial support, and there was nobody else to furnish capital we needed to complete the turn-around. We had no choice but to sell the business. We had been so frustratingly close!
SI: So what happened?Ferran: [Smiling] SubMicron was so deep in debt that there were no buyers; we couldn't give the company away! Then I got the idea of doing a management-led buyout. I went to a financial group that analyzed our market, talked to customers (who now said we were among their top-rated suppliers), and I was able to line up financing. This had to be done through an asset purchase after Chapter 11. I hope to never go through something like that again. It was too late to salvage any value for shareholders and bond holders, but there was a chance to take care of our employees, customers and suppliers. With the industry entering an upturn, I just couldn't give it all up. I'm particularly proud of the fact that, as Akrion, we voluntarily paid SubMicron's key suppliers and sales agents 100 cents on the dollar. We also took up all of SubMicron's commitments to customers.
SI: How did your people respond?Ferran: The people of our company are amazing. They've suffered so much, yet they never gave up. They are passionate about customer satisfaction and doing the right thing. Until recently, they either didn't have management support or the resources. Now they've both, and I'm confident they're going to shine.
SI: How does the future look?Ferran: At Akrion we say "It's a brand new day" and mean it. We've all learned from the mistakes of SubMicron. We've seen what happens when you put personal gain ahead of customer satisfaction. We understand that our job is to create value, not just for us but for all our stakeholders. With the strong financial backing we now have, and the industry on the upswing, I feel the future is very bright indeed. • -- Alexander E. Braun
Erratum
In the December 1999 article "1999 Editors' Choice Best Product Awards," the winning product line from Entegris was listed incorrectly as "Galtek GS" valves. The correct name is "Galtek SG" valves. •
 |
This SEM shows a photonic quasi-crystal lattice, formed in silicon nitride. Fabrication of the crystal is a team effort involving the research staff of the Southampton Microelectronics Clean-Room facility. The SEM bar marker is just 500 nm and shows the deep-submicron dimension of the holes and lattice. This crystal has 12deg of symmetry and exhibits a wide and isotropic photonic band gap. The high degree of symmetry allows band gaps to form in materials systems with a low refractive index contrast; this in turns allows integration with fiber devices/networks.
Please send submissions for SEM/TEM of the month to Peter Singer, Editor-in-Chief, 58 Summer St., Andover, MA 01810 USA. •
-- Laura Peters
