The Precision 5000 story
Applied Materials Inc., Santa Clara, California
Recognized for its process technology and design, the Precision 5000 family of CVD and etch systems is one of the semiconductor equipment industry`s most successful products. Chipmakers worldwide have purchased these systems for etch and chemical vapor deposition (CVD) applications. Introduced in April 1987, the Precision 5000 has just passed its 10th anniversary of production. The following article is excerpted from previously unpublished transcripts of historical material dealing with the company`s early product development activities.
The Precision 5000 was one of the industry`s first single-wafer, multichamber "cluster tools," and it was the first to achieve widespread industry acceptance. Its installed base now numbers more than 3500 systems. The Precision 5000 commercialized the single-wafer concept for plasma processing, and its multichamber architecture has served as the prototype for several generations of such tools.
The Precision 5000, with a central wafer-handling "platform" that could accept several small, single-wafer process chambers, was part of a forward-looking product roadmap envisioned by Dan Maydan, Sass Somekh, and David Wang. These three technologists, who had created Applied Materials` AME 8100 and 8300 etch systems, foresaw the need for complete system automation and exceptional flexibility that could be met with a "platform plus multiple-chamber" architecture.
Figure 1. Etch electromagnet coil configuration.
Jim Morgan: "We had a single-wafer effort in the late `70s, but it used one-at-a-time or single-station designs. We needed to get the throughput up. Using a multichamber design was a way to have excellent process control and get higher throughput.
"Once I saw our system, in the early 1980s - it was actually just a prototype on a bench - I knew it was worth backing. Chemical vapor deposition was a good potential market application. We had to have a product that wasn`t just equal to the other CVD technologies, it had to be a leapfrog technology.
"The key was to try to create a product shift. The 5000 CVD really created a leap forward in both process and design. It was a high-risk effort. I decided to bet all our resources on it because I believed in the design, and the processes were strong."
Dan Maydan: "Once we had decided to develop a CVD system, we were very clear about our requirements, which were that the process results were the most important thing, regardless of the architecture. But at the same time we felt the world was ready for a single-wafer CVD system.
"It was a major decision to develop single-wafer CVD as a multichamber system in order to improve productivity and process capabilities. We took an entirely new process direction where we needed a very high deposition rate, and we went to a new chemistry, using TEOS [tetraethylorthosilicate] instead of silane, because we felt much better step coverage would be needed going into the future."
David Wang: "When we started to work on CVD concepts, the industry was already running multilevel metal in microprocessors, SRAMs and power devices. Our target was to provide an economical advanced intermetal dielectric process for those devices.
"We spent several years working on various CVD concepts, both in configuration and in technology. We ended up with multiple chamber, single-wafer processing using TEOS for low-temperature applications, including the first self-cleaning in the industry. We also had the capability for integration, etch to CVD, or within CVD to produce multi-layer films. Also, we knew that the industry was going to 200 mm. The 200-mm wafer was becoming expensive, so batch CVD would be less attractive."
Sass Somekh: "In about 1983, I was working on an advanced single-wafer etch and David Wang was working on an advanced CVD system. I was also working on a handler to deal with multiple single-wafer chambers. Dan Maydan told me that we might be going to single-wafer and he wanted one mainframe that could be used for both etch and CVD. We started to work on what we called the 8400 in the summer of 1983, which was a multichamber concept that could take etch and CVD. The 8400 was the concept and feasibility phase that eventually led to the 5000 commercialization. The beauty of the 5000 was that the mainframe was designed at one time for CVD and etch chambers, so every square inch of the systems was thought through as to how it would work for etch and for CVD. And then, if somebody wanted to add a chamber, the chamber had to be designed to fit the system, and the system became a truly common module for all the chambers that went onto it."
Figure 2. Precision 5000 CVD system.
An aggressive schedule
Concept and feasibility work that led to the Precision 5000 began in 1983. Applied Materials formally kicked off its efforts to commercialize the 5000 on March 15, 1986. The schedule was very aggressive - the goal for the first shipment was December 1987, just 20 months away.
The Precision 5000 development program involved a much wider set of technical skills than were previously required in system and process development. The automation system alone required the company to develop its own wafer transfer technology, including robots, software, and controller electronics. Experts in process technology, automation, advanced materials, vacuum, heating control, particulates, and many other fields were needed to complete the project. More than 70 basic modules needed development, with more than 1000 custom designed parts.
Separating the "platform" development from that of the process and its associated chamber hardware had a considerable benefit in terms of development time. Once the platform had been developed, diverse chambers and processes for the many etch and deposition applications could be created quickly by teams focused on specific goals.
Dave Reis: "David Wang had a group of people making chambers in a back room of one of our old buildings. It was a grungy old building, but the group`s attitude was that it was the `skunk works` and that`s the way it should be.
"The 5000 CVD made a lot of sense. You needed some technical advantage to compete against the batch systems, and we had the plasma TEOS process that could do void free filling with excellent step coverage and improved deposition rate at low temperatures. But could we do it at a high enough rate? If you had multiple chambers, it made sense."
The race was on
Soon, however, a dilemma arose. A competitor had shown a new product aiming at the CVD market, and persistent rumors said that it was perhaps a year ahead of the 5000`s initial schedule. Jim Morgan and Dan Maydan met to discuss the issue. In a key decision, they decided to accelerate the final development effort to nine months from the original 20. The race was on.
The nine-month target proved impossible. However, in just 14 months a remarkable new technology was introduced, at a development cost of about $10 million (Figs. 1 and 2).
Dan Maydan: "From a project management point of view, the development was within budget, and we were able to give our customers their dream. We were giving them a single-wafer system with results which were better than any that were available at the time. But the decision we made that was quite tough, was to develop all parts of the product, in parallel, not sequentially, so we could come to the marketplace in a much shorter time. Toward the end of the commercialization phase, we had manufacturing, process development, and process chamber engineering, as well as platform engineering, all working at the same time."
David Wang: "Sometimes working in parallel presented a problem, but it is one way to accelerate the development. It is the best arrangement you can get to shorten the cycle time. We cherished innovation. We let every process engineer try new ideas."
Bang Nguyen: "It was fun to use the 5000 CVD from an R&D standpoint. A lot of the things we did were empirical - let`s try this and if it doesn`t work, let`s try something else. From the moment the decision was made to try something else - a new showerhead or a new susceptor - until we actually began experimenting with it in the lab, the time was relatively short.
"When the 5000 CVD came out, I worked for Applied Materials` main customer for the system. What we cared most about was whether you could run a wafer that met the requirements you were looking for, such as gap fill, or whatever. If you could do a few of those a day, you were in good shape. The first couple of machines we got were configured with all kinds of chambers. The single wafer concept was exciting. Before that, we were used to batch systems and furnaces for CVD. The 5000 CVD was good from a development standpoint because we were able to get wafers processed through a series of steps in a very short time."
Dave Reis :"When customers brought wafers into our lab, we could show them stuff that was magic. The problem was to try to duplicate it in a production environment. We were in the awkward position of not being able to build the systems and ship them fast enough to meet the demand, and at the same time we could not fix the problems fast enough and get those fixes out into the field because we were ramping the manufacturing so fast."
Sass Somekh: "These scientists suddenly were responsible for manufacturing. I really didn`t know much about manufacturing, so I read all these Japanese books about it. When we laid out the floor plan for the 5000 manufacturing area, we decided not to do a staging area where you bring all the parts from the stock room and you arrange them to go to the manufacturing floor. We adopted what they do at Toyota and all the big Japanese automobile manufacturing companies. By the time the 5000 went into manufacturing, the mainframe had only four parts in the stockroom and the rest was delivered to the production floor. That was a major accomplishment in advancing our manufacturing.
"I had some visitors from Japan who came over later on, and we showed them our manufacturing. They were impressed. I said, `Wait a minute, I read about how to do this in some Japanese books.` They said `Well, in Japan only the automobile companies do this, the rest of the companies don`t have this yet.` It turns out, that when one of our Japanese competitors wanted to impress visitors, they took them into their automated stockroom, but we had eliminated stockrooms altogether."
The Precision 5000 CVD system leapt from market entry to market leadership in under 12 months, won several awards and became the company`s most successful product introduction to that point.
The Precision 5000 CVD essentially introduced and commercialized plasma TEOS for multilevel metal applications. TEOS offers considerable benefits in step coverage and film quality over silane (SiH4), which was the industry standard at the time. Results with plasma TEOS were so dramatic that TEOS very rapidly became the industry standard chemistry for advanced CVD applications. Within about 18 months of the tool`s introduction, processes were made available for various doped and undoped dielectrics, as well as silicon nitride for final passivation.
David Wang: "The Precision 5000 changed the image of CVD. Before, etch was considered a very complicated, sophisticated process. But having the plasma TEOS intermetal dielectric application in a single-wafer design, with automatic self cleaning, suddenly made it a high-technology process that totally changed customers` perception of CVD."
Dave Reis: "Thermal TEOS was known to the industry`s technical people, but plasma TEOS was not well known. The plasma process was attractive because you didn`t need to operate it at 650 degrees. We were starting to go to multilevel devices. Intermetal dielectric was the big challenge at the time. Everyone was trying various schemes to planarize the surface so you could put another layer of metal on top. But it had to be done at a temperature that would not melt the previous aluminum layer. The Precision 5000`s plasma TEOS process results were so compelling that customers had to use it. We achieved high deposition rates at low temperature, with multiple CVD chambers for greater throughput, and also combined it with etchback for smooth planarization."
Kam Law: "I joined the company in September of 1984, and I spent about a year in concept and feasibility studies looking at TEOS for the intermetal dielectric. They asked me to check out whether the idea works, whether the step coverage was better than silane oxide. We started with a bell jar, basically a glass cylinder on top of a table, with a piece of aluminum metal as the base plate, and a piece of metal at the top. We tried different kinds of RF configurations, parallel plate, inductive coupling, and so forth. We just wrapped a coil around the tube on some of them. The data showed things that were very promising. We used the parallel plate idea, and we built a chamber, at first for 4- or 5-inch wafers, then we went up to 8 inch and scaled it down.
"TEOS was still kind of academic. LPCVD TEOS had been used in production at very high temperatures with a low growth rate. For PECVD TEOS, all the papers suggested that the film properties were poor. We spent a lot of time studying the fundamental physical and chemical properties in the lab and started to realize that, to our surprise, higher pressure optimized the TEOS oxygen flow such that it actually became more complete at higher pressure. In turn, you also got better step coverage and more complete oxidation at high pressure, and at the same time got the side benefits of higher deposition rate. Initially, when people heard we were going to deposit at maybe 8 or 10 torr, they laughed at us. PECVD at that time had always been below 1 torr, with low deposition rate and inconsistent film quality. Because of the high pressure, we could confine the deposition to the area we wanted, and at the same time we introduced in situ chamber cleaning, which was totally new to the industry. So everything was new, and it generated a lot of interest from people, but a lot of doubts as well."
Ted Iwasaki: "At an early stage, researchers in Japan had a problem with TEOS CVD. The (thermal) TEOS film had a tendency to crack. Applied Materials introduced the plasma-type process. Instead of ultraviolet or heat, it gave more energy to the process, which was more effective to dissociate the chemicals. Then customers were very pleased to see the conformality and geometry of TEOS CVD."
Kazuyoshi Yokota: "Japanese customers immediately noticed a dramatic reduction in particles and a big improvement in uniformity. The conventional system at that time was a tube LPCVD type of hot-wall reactor. They had lots of particles and extensive maintenance requirements. Five percent within-wafer uniformity was considered to be good at that time. With the 5000 CVD we were achieving 1 or 1.5%.
"The shift from single-metal to double-metal designs created a need for intermetal dielectric, and the 5000 CVD got into the market at the exact same time as double metal devices."
Chris Moran: "I remember watching the process development work on a prototype CVD chamber, which was totally manually controlled. To run a process, these guys would have to, with a good degree of timing and finesse, open valves, turn on gases, turn on power, turn off power, turn off gases, and so on. It was like ballet. They were dancing all over the place, flipping switches, turning knobs, and out would come these gorgeous wafers, but they did it entirely through their own skill. They would watch these big darkroom timers with the dayglo dials and flick the switches, then flick some more to go to the next recipe step. It was amazing to watch. They were designing the process chemistry on those prototype chambers for later transfer to more production-type machines."
Kam Law: "We spent a lot of time on the showerhead design. TEOS was very different from the silane process because the lifetime of the active species tended to be longer, compared to silane. If you provide uniform gas distribution and uniform pumping for TEOS, the gas coming out from the center has a longer period of time before it deposits, so the film is thin in the center and the edge will be very thick. It took a long time to figure this out, and we eventually had to force the gas toward the center before it got pumped out, to get uniform film deposition.
"The Precision 5000 also introduced the first adjustable electrode spacing for CVD. That turned out to be a powerful process parameter. The spacing between the active electrode and the ground where the wafer sits could be adjusted, which let us run more applications in the process chamber because, for example, different species have different lifetimes. Suddenly you had the option of using silicon nitride or silicon dioxide, or other processes, just by adjusting the spacing. Then everybody started to copy it.
"One significant breakthrough was when we got some device results from a really advanced customer. At that time, we didn`t really have the ability to characterize whether the film worked or not in the device. We could evaluate the physical and chemical film properties, and do the best we could to create good films with the best physical properties. At that time, in 1986, the customer was doing 0.8 micron. When they told us the film was good, device reliability was also good, we knew we had something for sure."
The Precision 5000`s original roadmap envisioned integrated sequential processing between chambers in one system under vacuum with "mix-and-match" process chambers. This "integrated processing" approach, however, turned out to be somewhat premature in the marketplace, as customers rushed to buy the 5000 CVD for its advanced process technology and productivity benefits, and overwhelmingly chose the "dedicated" configuration of multiple CVD chambers for maximum throughput.
Dan Maydan: "An important decision was the capability to put more than one process technology on the system, so we demonstrated etch and CVD very early in the game. From the beginning, the plan involved developing both CVD and etch chambers for the system, with the hope that some degree of integration - both types of chambers on one system - would prove attractive to customers."
Bang Nguyen: "There were some technical things that the system could do, like perform integrated sequential process steps between chambers under vacuum. But it was several years before our customers were ready for that. Today the concept is more acceptable than it was then."
Ted Iwasaki: "We talked about the advantage of integrating multiple processes. To the customers, integrated processing was a little too advanced at the time. They were reluctant to take a risk on integrated processes. They shared our opinion about its advantages, but they didn`t want to change their production tool sets."
Dan Maydan: "We penetrated the market at a rate that we ourselves did not anticipate. That was a catch-22. The more systems we sold, the less time we had to fix problems, and the more we needed to satisfy customers with new process capabilities, new applications. The fact that we had developed our own automation capabilities made us the only company in the world [at that time] to offer a viable multichamber system."
Almost a year to the day after the Precision 5000 CVD was introduced, Applied Materials announced the Precision 5000 Etch system. Using the Precision 5000 platform, the 5000 Etch was an advanced low-pressure, magnetically-enhanced reactive ion etch (MERIE) system.
Sass Somekh: "We were already working on the 8300 etcher when we realized we could use a magnetron concept to do a pretty good single-wafer etcher. We built a chamber that had permanent magnets in it, which we later changed to electromagnets. In the batch system, we could go to low pressure and get really good etching results. But when you wanted to etch in the single-wafer, you had to have an etch rate that was reasonably high and then the etch was not as good. The magnetron gave us the ability to determine what pressure range we wanted to operate in and still have pretty good etch rate for our single-wafer, because it made the plasma more intense and gave us a higher etch rate at lower pressure."
The Precision 5000 Etch system`s process library included a process for etching high aspect ratio single-crystal silicon with precise control, creating deep trenches and holes with smooth sidewalls and rounded bottoms. The advanced trench etch technology was instrumental in the development of a trench capacitor design for DRAMs.
Figure 3. Precision 5000 system with etch and strip chambers.
Figure 4. David N.K. Wang, Dan Maydan, and Sass Somekh pose with the Precision 5000 at its induction to the Smithsonian Institution in 1993.
Mei Chang: "David Wang said to me, `We need trench etch.` I did some process work on this huge development chamber we had. I made a process matrix and ran a cassette of 25 wafers. The condition was there. I knew what the behavior was going to be.
"But on the 5000 commercial product, the single-wafer chamber was a more compact design, and the process just would not work as well. I couldn`t figure out why. Finally I gave it to Jerry Wong, and said, `You go find out, I can`t make it work.` He came back saying, `It`s the materials.` We put quartz parts in the chamber and it worked. Alfred Mak took the chamber to the customer and he called back and said, `Oh by the way, the customer wants a trench like this, with two slopes.` I thought we needed some kind of sidewall deposition, so we added some deposition gas. In a couple of days we got the process in, went back to the customer and it was done."
Ted Iwasaki: "The 5000 Etch had a more difficult time than the CVD. There was a lot of competition. There also were strong connections between vendor and customer in Japan. In order to sell it in Japan, we had to try to cut the bond between the vendor and their customers."
Over the next two years, etch processes for polysilicon (October 1988), silicon dioxide (1989), and metal etch (October 1990) were introduced to the market on the Precision 5000 Etch.
Metal CVD launched
Launched in 1989, the Precision 5000 WCVD was Applied Materials` first metal CVD system, initially offering blanket tungsten deposition.
Mei Chang: "Most of our knowledge of the tungsten process came from work with selective tungsten, like how to nucleate on the substrate consistently. You can nucleate on metal or silicon or any conductor pretty easily, but not on insulators.
"We were thinking, `Tungsten is a metal, why does it look black? It should be shiny.` The tungsten surface was so rough, that`s why it looked black. So we tried preclean. We were like cooks trying different things to see which one come out better. One day I said, `Why not try using nitrogen?` People said, `Nitrogen? Nobody uses nitrogen.` I said, `That`s my point, nobody tried nitrogen before to see how it works, so let`s try it.` It was incredible. The surface that was rough before, now became flat. By adding nitrogen we got shiny tungsten; finally, shiny tungsten.
"Everybody was working in a similar temperature range, but pressure was a different story. We asked, `Why have we never worked in higher pressure?` In conventional LPCVD, everybody was pushing low pressure, but that didn`t make sense. So we forgot about LPCVD, just pushed the pressure higher. From 1 torr we went to almost 100 torr. We still use that pressure."
In May 1991, the process range of the system was expanded to include tungsten silicide. And in June 1993, a CVD TiN process was added to the 5000 WCVD`s advanced metallization capabilities; this process gives chipmakers the choice of either sputtered or CVD TiN films, depending on their specific requirements.
Within five years of the Precision 5000`s introduction, more than 50 unique processes were available on the system (Fig. 3). Its standard interface between the chamber and platform, defining robot movement and communications protocols, would also allow Precision 5000 chambers to be used by the company`s later-generation platforms, the Endura (1990) and Centura (1992).
Jim Morgan: "How do you stay ahead of the market? You have to hit the windows, otherwise you don`t have a product. You need to be sure your concepts are solid. You test them quickly. Then you resource it with really top-notch people. If you don`t have that then you don`t aggressively go forward."
In 1993, Dan Maydan, Sass Somekh, and David Wang were honored at the system`s induction to the Smithsonian Institution in Washington, DC (Fig. 4), where it is on permanent display as a key contributor to the creation of the Information Age.
For more information, contact Applied Materials, 3050 Bowers Ave., M/S 1263, Santa Clara, CA 95054.
The contributors to this article include James C. Morgan, chairman and CEO; Dan Maydan, president; David N. K. Wang and Sass Somekh, senior vice presidents and members of the Excecutive Committee; Tetsuo Iwasaki, Chairman, Applied Materials Japan, and president and CEO, Applied Komatsu Technology Inc.; Dave Reis, director of marketing, Installed Base Service and Support Division; Kazuyoshi Yokota, director, Global Business Development Group and CVD Product Business Group; Bang Nguyen, senior director of the SACVD Product Unit, CVD Product Business Group; Chris Moran, deputy general manager of the Metal Etch Product Group; Mei Chang, director, Advanced CVD Metals Product Unit, CVD Product Business Group; and Kam Law, general manager, Etch Product Division, Applied Komatsu Technology Inc.