Suppliers' successes with 300mm tools and materials
05/01/2001
Solid State Technology staff reports
overview
An industry-wide survey of some of the latest 300mm equipment and materials shows that the challenges of processing 300mm wafers are being understood and addressed. Many suppliers have devised clever solutions to the unique challenges in such areas as wafer reclaim and handling, lithography, gas handling, metrology, and wafer bumping.
It wasn't too long ago when experts in the industry were not sure who - certainly not any single company - would drive the development needed to get the industry to 300mm-wafer manufacturing. Historically, Intel led the conversion to 150mm and IBM to 200mm, both with the associated financial and technical pains. Paul Niemyski, process engineering manager at Texas Instruments' 300mm DMOS6,
Dallas, TX, notes that when the industry began to struggle with the transition, the operative question was "who would bear the cost of this wafer size conversion? We waited to see who would be first. Seemingly along with the rest of the industry, our strategy was to be the second or third manufacturer to transition to 300mm" (see "Managing 'process copy exact' . . ." on p. 109).
 Transistor module surface preparation area at DMOS6 at Texas Instruments in Dallas, TX. (Photo courtesy of Texas Instruments) |
Today, with the second, third, etc. 300mm fabs now coming on line (see table on p. 92), many fab managers are complimenting the supporting roles of this industry's suppliers of tools and materials. For example, Nun-Sian Tsai, senior director of the 300mm pilot line project at TSMC, Tainan, Taiwan, notes the "extremely supportive equipment suppliers" who have provided "valuable assistance during pilot line evaluation. Many were eager to install their systems in our Fab 6 pilot line since it would help determine which 300mm tools would be installed in our future plants, and those of the rest of the industry," says Tsai (see "Taking IC manufacturing from 300mm pilot to production" on p. 121).
300mm wafers
The story behind TI's 300mm DMOS6 fab start-up is particularly interesting. Here an engineering team used industry co-operation to drive 300mm tool specifications and unique internal methods to drive the development and qualification of 300mm tools and process integration, mainly using wafer processing capabilities at tool supplier qualification labs. In particular, TI's methodology was notable for its resourceful use of what at the time was "a very limited amount of quality 300mm diameter silicon wafers. The quality and availability of silicon required to do the evaluations that we planned posed a unique challenge," says Niemyski.
 300mm fabs coming on line, listed in order of operation start date, compiled by SST* |
Steve Brunkhorst in the 300mm business unit at wafer supplier MEMC, St. Louis, MO, tells Solid State Technology, that "'Are you ready?' is the 300mm question most asked by device manufacturers." This isn't just a question about geometry, rather one shrouded in increasingly difficult specifications for details such as surface localized light scatterers (LLS), particles, oxidation stacking faults (OSF), bulk iron, and other metals. "In addition, good reliability for some device designs depends upon the absence of bulk micro defects (BMD) in the device active region and increased perfection of the crystal lattice. Then, with increased crystal perfection, the elimination of COPs (crystal originated pits or agglomerated vacancy defects) from the device active region is of great interest to many device engineers," says Brunkhorst.
 300mm fabs coming on line, listed in order of operation start date, compiled by SST* (con't) |
Brunkhorst outlines today's three popular solutions for making 300mm wafers with adequate specifications of COPs. The first uses an elegant crystal growth process to suppress the agglomeration of vacancy and interstitial defects in the crystal bulk. While this delivers a very robust final product, development of the crystal process is a significant technical challenge. "Lack of certain demand for this product, combined with the technical challenge, means that COP-free grown crystals are not expected to be available in the marketplace from multiple suppliers in the near future," says Brunkhorst.
Another solution uses a high temperature wafer annealing process to repair agglomerated vacancy defects on the wafer surface and in the near-surface region. The principal technical challenge is the design of furnace hardware to preclude slip due to the interaction between high temperatures and gravity forces on the 300mm wafer. (Although a 300mm wafer has 2.25x more area than 200mm, it is only 7% thicker.) "The annealed-wafer product development is in advanced stages at some suppliers, but not yet commercially available," says Brunkhorst. The third solution is a wafer with an epitaxial silicon layer that covers COPs that intersect the surface of a "cost-optimized silicon substrate." The key technical challenge is a more straightforward scale-up of 200mm to 300mm epi-reactor technology; tool development has already progressed from advanced beta phase status to delivery of mass production suitable equipment. "Multiple wafer suppliers now offer epi wafer engineering samples as the first stage of product qualification," says Brunkhorst. "The epi family-of-products are receiving high scores from device manufacturers from diverse global locations, based on technical and business criteria. So, the answer is: 'Yes, we are ready, with epi.'"
A role for reclaimed wafers. With the limited supply of 300mm wafers, wafer-reclaim capability has also played a role in the development and transition to 300mm. Cliff Thomson, corporate technology manager at Exsil, Prescott, AZ, explains that reclaimed 300mm wafers have played a cost effective (Fig. 1) "bridge" role for tool marathon testing and particle monitors, and for applications work including thin film thickness measurements, oxide growth monitors, etch rate monitors, and implant dosing, among others.
"Reclaimers have improved their processes to provide reclaim wafers that meet prime test wafer quality requirements," says Thomson. "For example, even where there may be some reluctance to use reclaimed metal-layer wafers, we can provide metal analysis using graphite furnace atomic absorption spectroscopy." This means use of reclaim goes beyond just conventional cold processing applications to deposition, furnace, and ion implantation. "The percentage of reclaim can go as high as 70% and can decrease the total cost of test wafers by more than 50%," says Thomson.
300mm wafer handling. Other, bigger issues for suppliers of wafer handling solutions have arisen. Michael Wright, president of Entegris's Microelectronics Group, Chaska, MN, notes it is crucial to understand the importance of interoperability and interchangeability in 300mm fabs. "With more than 500 process steps in the modern 300mm fab, we have the challenge of '1 in 5 million.' At a minimum, there are 5 million interactions with a FOUP as it passes through a fab. If any one of these interactions fails, because of wafer plane alignment for example, the fab process flows are negatively impacted," he says.
As demanding as it is, FOUP technology is one area where 300mm fab technology is relatively mature. Entegris, for example, gained a great deal of knowledge by working with every company with a 300mm program, developed the very first FOUP, and is now in its fifth generation of product learning.
 Figure 1: Cost savings from using 300mm reclaim wafers is substantial when prices are compared to virgin monitor wafers. (Source: Exsil Inc.) |
Wright says, "The value of our customers' wafers is so high that any issue such as abrasion, contamination, and predictability must be addressed long before inserting our products into the fab." According to Wright, Entegris is fortunate to have worked with the first 300mm pilot and production lines in the world. "Our experience with automation and platforms from every equipment and automation supplier in the industry has heightened our awareness of the many different interactions between systems and our product. We are the common denominator in 300mm. We have to have the right materials, structure, system, and predictability for all the other systems in the fab to work," he says.
Automation and connectivity
Indeed, the intensity behind FOUP automation extends throughout an entire 300mm fab's level of automation. Paul Sagues, president of Berkeley Process Control, Richmond, CA, says, "The control system has to be rock solid across a 300mm fab and capable of recovering from errors. With 200mm, we had the fallback of carrying around cassettes of wafers and reaching into machines with a vacuum wand to unload one that needed to be re-booted. But with 300mm, we have to rely on automation."
With 300mm fab automation, the industry sees the need for hardware standards and inter-operative software; the patchwork approach of proprietary software is not a good choice. For example, Berkeley Process Control has adopted Ethernet, now extended down to 100Mbit, as its control network rather than one of the two dozen proprietary device-level networks. "We focus on lower-level machine operations, and others are working closer to the enterprise level. Clearly, Ethernet is the right hardware link. In the future, various software connectivity standards - all based ultimately upon TCP/IP protocols - will evolve, but it's too early to say whether it will be OPC, CORBA, HSMS, OLE or some other communications standard," says Sagues.
Process control challenges
Sagues believes that there is confusion where the enterprise meets the real-time level. "Some are putting real-time operating systems on the NT platform. But the NT platform has to be free to evolve; putting a real-time operating system on it ties the NT platform to the hardware. And the other extreme - putting TCP/IP connectivity at the sensor level - will never fly because it is trying to put enterprise connectivity into the control system," he says. Sagues believes the control requirements associated with 300mm means getting data out of the high performance control system at a sufficient bandwidth that the upper-level optimization software can do its work. "Clearly, the high levels of advanced process control should not be done on the control system, but without a seamless integration between the control system and the upper-level advanced process control computers, the system will not yield results."
Lithography
While the development of 300mm tools has not been a straightforward proposition at any process step, coming up with the right lithography tools has been particularly challenging. The industry's ramp to 300mm wafers has occurred in a period when lithography system suppliers have also had to develop bridge tools and extend optical lithography to 193nm (ArF source) capability, not to mention the need to closely watch the emergence of next generation lithography technology.
 Figure 2. An example of a "down-sized" sensor is the new Micro-Baratron from MKS Instruments. |
Consider Canon, for example. Since 1995, Canon has evolved three generations of 300mm tools - EX3L and i5L steppers installed in 1997-98 at Selete, I300I, SC300, among others; the 200mm-300mm convertible 5000 scanning platform used to conduct many 300mm wafer demonstrations to ensure production capabilities; and, in the fall of 1999, a 300mm-capable FPA-5000AS1 ArF scanner installed at Selete. By 2000, when most 300mm fabs were still shells, Canon was into its third generation of 300mm lithography tools capable of mix-and-match lithography down to 110nm.
Canon's efforts, like those of other suppliers, have also been with bricks and mortar. The company is expanding its manufacturing complex in Utsunomiya, Japan, spending more than $260 million to address 300mm tool commercialization. It has also built a new headquarters and research facility in the heart of Silicon Valley, where its objective is to "readily coordinate work with other suppliers who have application development facilities in place for processing and evaluation, especially for 300mm wafers, beyond the lithography sector," says Joyce Heuman, strategic marketing manager at Canon USA, Irving, TX.
Indeed, Canon's efforts with 300mm technology are a clear example of how semiconductor equipment and materials suppliers have funded the industry's 300mm transition. Heuman notes, "Our San Jose, CA, applications lab and development program helps customers work out processing details and fine-tune equipment integration very early in the rush to a new product generation. And they can work out a turnkey solution for 300mm wafer processing, not just the most cost effective tool set mix, but a complete recipe with resist, exposure and etching parameters for each layer."
Gases for 300mm manufacturing
One might assume that the transition to 300mm wafers would be relatively straightforward for suppliers of gases and gas handling equipment. However, Benjamin Hertzler, technology development manager for the Electronics Division of Air Products and Chemicals Inc., Allentown, PA, explains "This transition is having significant implications for manufacturers of process gases and delivery equipment. For example, specialty gas flow rates for 300mm fabs are expected to increase two to three times those for 200mm." He predicts that this increase may require a dedicated cabinet/tool. In full-scale 300mm fabs, large volume gases such as NF3, N2O, NH3, SiH4, and HCl are best supplied using bulk specialty gas supply (BSGS) systems. In some cases, additional supply considerations, like heated cylinders and gas heaters, are needed. "Although the piping system's total length is not expected to increase significantly, the diameter for certain specialty gases will require half-inch lines versus traditional one-quarter lines commonplace with BSGS systems," he says.
"Gas systems will also have to be more automated and interface to host computers simply because 300mm fabs will be automated," says Hertzler. "In addition, because of increased capital investments for 300mm factories, equipment reliability will be more closely scrutinized. A supplier's experience plus operating data will be invaluable. The migration to 300mm fabs will be accompanied by an increase in process complexity. As a result, device manufacturers will outsource more fab operations as they concentrate on device design and process technology. An option is to offer BSGS as an own, operate, and maintain product supply agreement provided by the gas/equipment supplier to foster reliability."
 Figure 3. Semitool's Spectrum 300 300mm automated, batch wafer chemical processor, for stripping, critical cleaning, and etching applications. |
Gas delivery at the tool level is also an important part of 300mm fabs. Here a driving factor is tool footprint, which is typically required to be no larger, and in some cases smaller, than the equivalent 200mm tools. Allen Hood, a product manager at MKS Instruments says, "For many tools that use process gases, a large amount of space is required for gas distribution systems. To save space, the industry is moving away from gas supply systems connected with in-line VCR fittings to more compact "gas sticks" with surface-mount fittings. Using this streamlined design, individual components can be closely spaced yet easily removed and replaced without disassembling the gas line. In turn, gas stick components must be small enough to match tight spacing requirements and to facilitate servicing. Components that send output information should provide a local visual pressure display on or near the process tool." (Fig. 2).
Process equipment flexibility
Equipment suppliers are keenly aware of 300mm tool cost after going through an industry-wide false start a few years ago. One way to limit the cost, at least for a company that makes more than one type of tool, is to use a common platform for multiple tool types. Trikon Technologies, Newport, UK, has done this with its physical vapor deposition (PVD), chemical vapor deposition (CVD), and etch equipment. Trikon used a Brooks Automation Gx8000 platform for that set of tools to maximize commonality. Existing controls and software from previous tools were implemented to reduce development cost.
Another common strategy is the use of "bridge tools" that offer flexibility by accommodating 200mm and 300mm wafers with relatively simple modifications. This offers the user more options and the ability, for example, to switch a 200mm fab to 300mm more easily, or to develop the processes on 200mm equipment so that fewer variables are being changed at a time.
 Figure 4. The spin speed for application of SiLK should be 1000-2000rpm on 300mm wafers, according to data from Dow Chemical. |
For 300mm etching equipment, the ITRS Roadmap predicts high-density plasmas for all applications beyond the 100nm-technology node. (High-density plasmas are currently used for metal and polysilicon, but medium-density plasmas are currently used for dielectrics, including low-k). When scaling up etch applications to 300mm, there are a number of important parameters to control. To address these upcoming needs, Trikon's M0RI source has inherently good uniformity, and a system of electromagnets helps to control the plasma and further improve the uniformity. This is more important as the wafer size increases. The capabilities demonstrated include 25nm vias and 30:1 aspect ratios for oxide.
Another bit of flexibility in Trikon's set of tools is CVD equipment that can be used for dielectric gap-fill in aluminum processes as well as dielectric layer deposition in copper damascene processes. This allows the user to make the switches from 200mm to 300mm and Al to Cu independently.
"This is a serious consideration for DRAM and SOC companies since they will be making the move to 300mm as early as possible to reduce costs," reports Bernard Culverhouse, VP of marketing at Trikon, describing one source of the motivation for that flexibility. "Initially, this will be with aluminum technology for all design nodes down to 0.1µm, and thereafter, with copper."
Tool flexibility is also featured in a system from Semitool, Kalispell, MT, whose Spectrum 300 product (Fig. 3) is 200mm- and 300mm-capable, with process scalability from the 200mm version. It is an automated batch wafer chemical processor for stripping, cleaning, and etching applications. It was also designed to have low chemical and DI water consumption, low exhaust, and a small footprint. These environmental factors are increasingly important for the most modern fabs.
Equipment components
Much of the challenge with 300mm wafers is associated with the wafer's physical size, and this is especially true in a process like CMP that has a significant amount of motion as part of the process. Kollmorgen, Radford, VA, a supplier of components to process equipment manufacturers, has a "direct drive rotary" (DDR) technology that addresses the mechanical challenges of a 300mm CMP process.
A CMP tool typically has multiple heads and platens that drive the polishing of the wafer. The DDR approach eliminates gearboxes and timing belts by coupling the load directly to the servo motor shaft. "This is achieved through the use of special magnetic design torque motors, typically large diameter, short axial length motors with a very high number of magnetic poles to achieve the highest torque per volume," according to Tom England, director of product management at Kollmorgen. "This is particularly important in 300mm machines since the required machining torque is "considerably higher than that of the 200mm counterpart.
Kollmorgen's streamlined approach is also in line with many suppliers' efforts at reducing the part count in complex equipment. The reliability is also likely to improve, because removal of a gearbox eliminates oil or grease leakage problems, and removal of timing belts reduces particle generation.
Processes and materials
Process uniformity on 300mm wafers is a challenge for the material suppliers as well. With spin-on dielectrics, for example, process parameters such as spin speed, solvent evaporation, and cure times can have a significant effect on film uniformity and thickness. Not only the process but also the material itself must be designed for 300mm processing.
Dow Chemical, Midland, MI, reports that it has been working with spin-track tool suppliers on these processes using new formulations of its SiLK dielectric resin to address the challenges of 300mm spin processing. Progress to date includes information about the spin speed process window (Fig. 4). 200mm tracks typically work best at 2000-4000rpm, but 1000-2000rpm is better for 300mm. The slower spin speed increases the film uniformity challenge.
Sputtering targets present other kinds of challenges for 300mm tools (Fig. 5a). With the larger target diameter comes the challenge of maintaining a structurally stable sputtering target assembly that will not deflect or bow under normal operating conditions inside the process chamber. MRC, a division of Praxair Surface Technologies, Orangeburg, NY, used modeling to design its target assemblies to address this (Fig. 5b).
MRC thinks the overall strength of the assembly is critical with such large structures. Its investigations have shown that a bonded assembly exhibits higher strength characteristics than single piece construction. MRC has developed a bonding technology for attaching high purity targets to backing plates that results in an improved interface and does not affect the fine microstructure of the target.
Metrology concerns for 300mm
The greater value of 300mm wafers and the equipment processing them increases the importance of metrology. A bad wafer or a tool functioning out of its control limits is a bigger liability than before. Even the investment community has realized the critical nature of metrology capabilities during the transition to 300mm wafers, with metrology tool providers getting some better investment ratings than the process equipment makers during the industry downturn.
Shallow trench isolation (STI) will be an important technology for several device generations, so characterization of shallow trench systems on 300mm wafers is a valuable capability. Traditionally, this has been accomplished through the use of time-consuming and highly localized techniques like atomic force microscopy (AFM) or through the use of destructive methods like scanning electron microscopy (SEM). While capable of yielding detailed information regarding the nature of such systems, these techniques do not generally accommodate high-throughput sampling. As a result, they do not lend themselves to wafer uniformity studies and cannot practically be used in volume production or manufacturing environments. In addition, the transition to a 300mm fab means that these traditional methods will require even more time to produce usable wafer characterization data.
n&k Technology, Santa Clara, CA, has developed a nondestructive method of characterizing shallow trench structures, capable of mapping an entire 300mm wafer in minutes. This method is nonlocalized, returning average results from an illuminated measurement area of 50µm. In addition to measuring standard quantities like trench depth and width, the system is also capable of acquiring data about the thicknesses and properties of films deposited inside and outside trench regions.
In a test comparing the n&k 3000 Analyzer to SEM trench measurements, the average trench depths agreed very well, but the n&k tool also revealed center-to-edge variations and trench geometry process variations on the order of 100Å. This is the kind of data that would be difficult and time-consuming with SEM or other traditional techniques, especially as the wafer size increases.
Another metrology technique available in a tool targeting 300mm wafers is total reflection x-ray fluorescence (TXRF), which is used for metal surface contamination analysis. Atomika Instruments, Tempe, AZ, has introduced a 300mm full wafer TXRF tool that provides simultaneous quantitative analysis of most relevant elements used in semiconductor manufacturing. Like n&k's tool, it is nondestructive, which helps the industry's drive toward reducing the quantity of test wafers required.
Atomika's TXRF 8300W incorporates the GEM interface with SECSII protocol, and implements SEMI E58 Automated Reliability, Availability, and Maintainability Standards (ARAMS), which allows equipment to send utilization, reliability, and maintainability information to host systems using a set of predefined messages and a standard performance model. Its integrated Class 1 mini-environment is suited for a Class 1000 production environment. The tool was qualified for 300mm production at the Dresden SC300 facility.
In situ analysis is identified in the ITRS as a key metrology technology for meeting upcoming milestones. On-line, real-time information about tool and product status is needed to increase productivity and quality.
 |
Inficon, Syracuse, NY, has created a system that addresses this and has been installed in 300mm PVD tools. Its FabGuard Integration and Analysis System provides quick, automatic feedback on processes and equipment conditions, allowing the process engineers to spot problems sooner. This system of sensors and software has successfully reduced the number of lost wafers due to inborn contamination from upstream processes, according to Steven Hisel, process metrology marketing manager at Inficon.
 Figure 5. a) A sputtering target, and b) modeling results showing deflection of a Ta on Cu/Cr baseplate target assembly at 100°C. (Source: MRC) |
The in situ system allows process engineers to gather sensor data from multiple sources, such as residual gas analyzers and particle detectors, and also directly from the semiconductor processing tool. Data analysis routines, both real-time and run-by-run, can determine if key process parameters are out of spec, and then notify the responsible people for a timely response. The system can also be used to "fingerprint" a process chamber after a cleaning process to make sure that it is ready for production. Similarly, it can be used to determine when preventive maintenance is needed and when it isn't, thus preventing unneeded maintenance on a good tool, or processing on an unfit tool. As equipment and wafers become more costly, the importance of these in situ capabilities increases.
Wafer bumping and probing
A new factor in the current change in wafer size is the effect it has on packaging processes. With wafer-level packaging emerging, many of the assembly and test equipment makers now have to take wafer size into account.
"The advent of 300mm wafer processing poses new challenges for bump technology just as it does for all other levels of semiconductor manufacturing," states Stephen Kay, director of product marketing for packaging technology at Ultratech Stepper, San Jose, CA. "The critical driver in a successful transition to 300mm will be achieving acceptable levels of yield. We believe that shifting to 1x wafer steppers for bump lithography applications provides the most technologically and economically feasible alternative to traditional contact/proximity aligners."
Ultratech has had some success with its approach to 300mm wafer bumping, by, for example, collaborating with Flip Chip Technologies, Phoenix, AZ, and others to produce the first 300mm bumped wafer last year (Fig. 6). Ultratech sees several things driving the 300mm wafer bumping market. The obvious ones are the industry-wide shift to 300mm wafers, the growth of flip-chip and wafer-level packaging, and the increasing economies of scale for wafer bumping at 300mm.
Another factor is the types of devices driving the move to 300mm wafers. The technology of packaging or bumping ICs in wafer form is applicable to a wide range of devices. Low I/O devices such as integrated passive devices or EEPROMs rely on wafer level CSP technology for a small form factor, low-cost packaging solution. High performance, high I/O devices such as microprocessors and peripheral logic use flip chip mounting not only for its superior electrical performance and low parasitics, but also for its ability to enable pad-limited designs that cannot be wire bonded. Fabrication facilities are being built around the world to produce a number of different devices on 300mm wafers, including DRAM, logic, and microprocessors. These high performance devices will be the first products to be bumped on 300mm wafers. As the technology and availability of 300mm processing broadens, more cost-sensitive devices for consumer products will begin to be produced on the larger size wafers.
Wafer probing is another back-end function having to make changes to deal with 300mm. The Micromanipulator Company, Carson City, NV, a manufacturer of analytical probing stations, is overhauling much of its approach to address the challenges of probing larger wafers.
"From 100 to 150 to 200mm, we have been able to 'scale-up' products," says Michael S. Jackson, director of sales and marketing at Micromanipulator. "The 300mm transition has caused major product redesigns instead." Jackson described several things they are developing, including:
- 300mm probe stations with motor controls even when used in a "manual-interactive" mode. The size of the tool prevents the user from reaching controls that are at its rear, and it is not practical to manually rotate a knob to move with 1µm or less resolution across a 300mm distance to probe the wafer edges.
- Analytical probe stations with wafer handling capability, instead of manual handing feasible with smaller wafers. With 300mm wafers, it is becoming common for customers to request this. "Users just do not like the idea of picking up a 300mm wafer with a vacuum wand and inserting it onto a chuck in a probe station," said Jackson.
- A redesigned probing system to meet the tighter positioning requirements for 300mm wafers. The 300mm tool has greater travel than the 200mm, but position error has been decreased from 1.5µm to 1.0µm.
- 300mm probe station with accessories from its 200mm probe stations, allowing analysis work in facilities with bridge tools or a combination of 200- and 300mm processes.
Jackson has also seen some encouraging signs from users, in spite of the current slow-down in the industry. He has seen device manufacturers who are continuing with 300mm plans because they have learned that they can't turn their suppliers - the equipment makers - off and then on again as quickly as would be needed when the demand increases again.
If this is a real trend, then the industry should be in good shape to continue the 300mm progress described here.