WAFER HANDLING AND AUTOMATION

Controlling wafer transport in a vacuum environment

Martin P. Aalund, Smart Machines

The construction of a semiconductor wafer fab today is estimated to cost $1 billion to $1.5 billion, and price tags in the $2-billion-plus range are rapidly approaching. These floating fabs—literally concrete boats floating inside a foundation—reduce contamination to Class 1 levels at the workspace. According to Dan Hutcheson at VLSI Research (San Jose), the semiconductor industry is building about 25 fabs a year. At such a high cost, however, it's difficult to justify the economics.¹

Process tools within the fab cost millions of dollars. As a result, each process tool must contribute to a profitable return on assets. Nonproductive downtime caused by tool failure or malfunction has a substantial impact on the bottom line. With a minimum of 4 hours to pump down to vacuum after a failure or malfunction, an extraordinary loss of revenue occurs. Considering the fact that large fabs have up to 200 process tools, the numbers are overwhelming.

Not only are the costs associated with a fab and process tools staggering, but the value of the product on each individual wafer also falls into the same category. The Semiconductor Industry Association predicts that circuit density will increase by a factor of 20 by 2010.² An increase in density, coupled with an increase in area of two times, from 200 to 300 mm (or potentially four times with backside processing), could result in up to 80 times as many components per wafer. While component complexity will increase over time, a 300-mm wafer could be worth between two and 80 times its 200-mm counterpart.

The processing demands of 300-mm wafers are real today. Industry timetables call for pilot 300-mm factories to be on-line by the end of 1999, with some production fabs expected to be operating by 2000 or 2001. The transition to 300-mm wafers is estimated to represent at least a $14-billion opportunity for equipment and materials companies. One benefit of switching to 300-mm wafers is a decrease in device manufacturing costs by as much as 20 to 30% while producing more than twice as many chips as is now possible with 8-in. wafers. The elimination of human intervention in the manufacturing process is another benefit.

Figure 1. With wafer fab costs breaking the $2-billion mark by the year 2000, productivity is mandatory. (Source: Paine Webber.)

Productivity is mandatory with $1 billion to $2 billion fabs and is particularly important given rapid equipment devaluation. This spiraling cost is expected to continue as shown in Figure 1. It is common to write off a fab in three years in the United States, which translates into a depreciation rate of $1600 per minute, or $96,000 per hour. Automation content is approximately 1% of the fab cost, or $15 million to $20 million (see Table I).³ With 300-mm production, the industry requires a more controlled process. There also is a growing need for more-sophisticated tools that use feedback loops for real-time monitoring of processes such as pressure in the chamber and the igniting of plasma for etching at a precise moment.⁴ Although vacuum technology has been in existence for a long time, recent and continuing advances in vacuum process tools make a vacuum environment a reasonable alternative to constructing new fabs to achieve similar goals.

Equipment Types	Typical Units	Average Selling Price ($ Million)	Total ($ Million)
Chemical vapor deposition Physical vapor deposition Total deposition	24 23	$2.5 $3.5	$60 $81 $141
Exposure tools Photoresist processing Total lithography	54 54	$6.5 $2.0	$351 $108 $459
Etch Cleaning/strip Chemical mechanical polishing Total etch, clean, CMP	55 30 20	$3.4 $0.6 $1.2	$184 $18 $24 $226
Process diagnostics/Metrology Diffusion furnace/RTP Ion implant Factory automation Miscellaneous	N/A 32 13 N/A N/A	— $1.0 $3.3 — —	$60 $32 $42 $15 $67
Total front-end equipment Land, building, and facilities			$1042 $521
Total			$1563

Table I: Capital spending for a 0.25-µm, 200-mm fab.³

History of Vacuum Technology

As fab construction costs escalated during the mid-1980s, it was recognized that there may be an advantage to miniaturizing cleanrooms by producing products in a vacuum and transporting wafers using robots (see Figure 2). For a while, boutique fabs stimulated substantial vacuum technology development. Then later, when large fabs again replaced boutiques, new developments were easily adopted by large manufacturing facilities. During the late-1980s, a transition from 150- to 200-mm wafers began. Although 80% of semiconductors are manufactured from 200-mm wafers, it has taken 10 years to approach these levels.

Figure 2: State-of-the-art fabs use robots to handle wafers in both atmosphere and vacuum. Buffer stations are used on the front end of process and metrology tools to ensure full tool utilization.

Today, two discontinuities are taking place. The first is a move from 200- to 300-mm wafers that will almost double the surface area of each wafer. Microprocessor and memory manufacturers that process 30,000 wafers monthly will process twice as many products with 300-mm surfaces in the same amount of time. The second development involves process geometries. In 1994, 1-µm technology began to move to 0.5 µm and then 0.35 µm, which today accounts for 50% of the market.

Migration toward 0.25- and 0.18-µm technology is under way. This trend is significant for two reasons: 0.25- and 0.18-µm devices are substantially smaller and more difficult to build, and their manufacture pushes the existing envelope of optical equipment technology. The products also may become I/O bound. New, higher-density packages will need to be developed. Although substantial research is exploring possible solutions including x-ray and E-Beam projection systems, no major breakthrough is imminent. What is slowing the move to 0.18- and 0.25-µm technology is that process equipment is not yet available. VLSI Research estimates it will take the market until the end of the next decade to arrive at 80% of wafers at 300 mm and 0.18 or 0.25 µm, with the remaining 20% comprising 150- and 200-mm wafers.¹

As the market embraces 300-mm capability, safety, direct-drive vacuum robotics, and communications standards are becoming critical issues. In addition, 300-mm applications may be further complicated by backside processing, which will affect the contact area for wafer handling—the outer 1 to 3 mm. Prealignment, or centering, is necessary to prevent incursion into the defined exclusion zone and standard errors in alignment of a wafer in a cassette or elevator. Today, most passive-bladed end-effectors are not up to this challenge. New edge-gripping technologies will be needed to meet increased throughput demands of the wafer transport system while maintaining a contaminant-free environment.

The shift to 300-mm wafers requires manufacturers to purchase new equipment that must be fully automated. With 300-mm wafers, which each weigh 0.35 lb versus 0.14 lb for 200-mm wafers, full cassettes or pods are too heavy for repetitive handling by humans. The added value of product per wafer favors an automated solution as well. At 0.35 µm, it is still possible to process semiconductor wafers in a Class 1 cleanroom in an open fab with atmospheric conditions. As the industry moves below 0.35 µm, the recommended method to ensure noncontamination is to process wafers completely in a vacuum environment, which ensures that airborne contaminants are eliminated.

Figure 3. As these tests for a vacuum robot show, particle generation must be minimized or yields can be jeopardized.

With a vacuum environment, several contamination problems are solved, but new challenges regarding the survival of mechanical parts and electronics within a vacuum are created. Particulate generation must be minimized; particles cannot remain suspended in air columns where they may potentially float down onto a wafer. With today's new technology, robots generate particles at a much lower level than the Class 1 limit. Figure 3 shows the particulate generation for an advanced vacuum robot.⁵ Particles were measured at a variety of positions around the robot while it was in motion. Even the worst-case positions are far below Class 1 requirements.

Within a vacuum, reliability, cleanliness, and dependability are paramount. Furthermore, the system must be sealed to maintain a vacuum. The sealing system design must allow for movement, yet be durable and reliable enough to withstand millions of cycles. Lubrication systems and materials must also be specific to the vacuum environment. During the past 10 years, substantial advances have been made in vacuum technology and in improving the operation of robots within a vacuum. This technology will be further advanced to meet the challenges of 300-mm manufacturing.

Technological Advancements

Until recently, the semiconductor manufacturing process suffered from wafers breaking in vacuum process chambers whenever power failures or emergency stops occurred. This breakage could become more frequent with the larger 300-mm wafers. Recent technological improvements in some robots, however, ensure that during an emergency stop, the robot arm and drive will not damage the robot or wafers in transport. The arm decelerates at a safe speed along the path it traversed prior to the outage. This feature may be mandatory for all robotics systems used in 300-mm processes.

Fewer particles in a vacuum environment mean increased die yield in 300-mm deep-submicron semiconductor processes. Since bearings generate particles, reducing the number of bearings in a vacuum and enclosing them substantially reduces contamination. Although some designs still have as many as 10 bearing sets in a vacuum, which increases the probability of microcontamination, other vendors have been able to reduce bearing sets to as few as two.

Mechanical. Recent advances in direct-drive architecture versus gears and timing belts have led to a more than 10 times improvement in reliability. New-generation robots are capable of greater than 10 million cycles between failures compared to conventional robot architectures which can complete fewer than 1 million cycles. New-generation robots also provide greater positional repeatability and smoother movement, which ensures increased acceleration of wafer movement, and does not contribute to vibration. Direct-drive designs that attach motors directly to a driveshaft further minimize the number of moving parts that are subject to mechanical wear, and therefore provide extremely high reliability and optimize wafer throughput.

Materials. Specific materials are necessary in vacuum environments and in applications exposed to process chemicals. Gold plating, for example, is optimal for exposed wiring, since copper quickly corrodes when exposed to trace amounts of some process chemicals. Trace gases cause unexpected arcing on even modest voltages during pumpdown of vacuum chambers with potential product loss. All mechanical and electrical components in a vacuum must be capable of surviving trace quantities of chemicals used inside the process chamber while being compatible with a vacuum environment. For example, copper and other high-valence metals are good conductors in electrical systems but will outgas in a vacuum. Furthermore, copper quickly dissolves when fluorine, chlorine, and other corrosive chemicals are present in processes. Arm and blade materials must be suitable for specific processes as well. In some cases, blades must be ceramic, and stainless steel must be shielded from direct chemical contact.

Electrical. Vacuum processing is not without difficulties. One of the most complex challenges of electronic components in a vacuum is the connection between integrated circuits and PCBs, and associated wiring problems. Simple solder connections must be encapsulated in vacuum-compatible compounds since common solder will outgas in a vacuum, creating an additional source of contamination and component failure.

Figure 4. Electrical wiring that is exposed to vacuum must be designed to eliminate virtual leaks and withstand harsh conditions.

When standard wire is exposed to a vacuum, not only can the insulation outgas, but gases trapped between the airtight insulation and wire, or in a multistranded conductor, may act as virtual leaks, or create bulges in the insulation as a vacuum is drawn (see Figure 4). In some instances, this results in the delamination of flex circuits or rupture of wiring insulation. Jacket material must be manufactured from vacuum-compatible materials such as Kapton (3M, Austin, TX). Connection between wires must be created by pressure contacts or with nonoutgassing solder, or by embedding solder joints in a vacuum-compatible epoxy.

Within a vacuum environment the cooling of electronic and control equipment is also challenging. Areas of heat generation include power supplies, power electronics, and control electronics. High clock rates on cpus generally produce high thermal loads. A good design must include substantial emphasis on heat dissipation in an enclosed environment. Even though electronics may be capable of operating at 100°C, elevated temperatures generate high stresses on bearings and mechanical components because of thermal expansion. Ultimately, these high stresses on bearings may lead to physical damage or grease breakdown and result in premature failure.

Figure 5. Resolvers use magnetic fields to provide robust absolute angular position information.

If brushless resolvers are used, an electromechanical homing flag system is not needed, and system reliability is improved. A resolver is a robust electromechanical position sensor that has a long history in robotics and control applications (see Figure 5). The device operates as a transformer with coupling that varies as a function of its rotor position. An input carrier frequency is fed into the device, and two modulated outputs of the same frequency are returned. These two output signals can be used to determine the absolute position of the robot.

An advantage in recently announced through-the-arm wiring technology is that it allows for placement of sensors on the end of the robot, with active grippers on the end of the arm to further increase speed and throughput. This wiring provides a pipeline for information and power to the gripper, so that robots can sense gripping and perform other functions effectively, with no barrier to power. The robot now has "eyes" to provide advanced sensing and "hands" to safely grip the wafer during high-speed transport.

Semiconductor devices and PCBs require special handling as well. These devices must be protected from the environment while providing adequate thermal paths to dissipate heat. To date, there are three methods of providing this protection:

• Enclosing: Devices are sealed in a container at atmospheric pressure. This shields the vacuum from outgassing from the device and protects the device from effects of the vacuum. This method solves both contamination and environmental issues but requires that pressure must be constantly monitored to ensure no leakage is occurring. Even slow leaks will reduce internal pressure enough to degrade performance of the circuit or contaminate the chamber. Pressure sensors should be used to monitor the enclosures.

• Potting: Devices are completely covered with a compound such as a vacuum-compatible epoxy to minimize outgassing and alleviate mechanical stresses. This method can greatly reduce the effects of vacuum on electronic devices. Care must be taken to minimize the area of the exposed potting compound by placing the circuit in a vacuum-compatible recess or lidded container, and filling the inside with potting. Potting eliminates problems associated with leaks; however, the whole assembly must be scrapped if any one component fails.

• Venting: A path is provided to the atmosphere for a sealed chamber, allowing the device to operate at atmospheric pressure. This method has many of the same advantages of enclosing; however, slow leaks can be tolerated as long as they do not compromise chamber cleanliness.

New European Community requirements place demanding constraints on all electromechanical systems destined for Europe. The system must meet both electromagnetic susceptibility and emissions standards stated in the electromagnetic compatibility directive.^6—8 In addition, automation may need to meet both the low voltage and the machinery directive.

Controls. Software plays a major role as complex algorithms cut prep times, enhance the speed of wafer handling, and provide for smooth material flow. By using sophisticated motion control software algorithms that were designed for a direct-drive system, wafers can move at the highest possible speed without slipping. In addition, continuous rotation without having to reverse direction can be accomplished. Already, in some situations, the results are 100 wafers/hr versus 30 wafers/hr; the improvement is mostly attributable to software. These control features result in a lower cost-of-ownership for process tools, greater throughput, and better wafer yield. By monitoring wafer accelerations, the system can prevent wafer slippage during a move. An additional benefit of advanced software control is that systems are readily optimized and reconfigurable to new customer requirements. Current software anticipates all wafer movement, paving the way for self-correcting machines.

A key to optimizing chamber usage is reducing cycle times, or the time it takes to move a wafer from one process chamber to another. One of the most common methods of movement is the use of multiple blades that allow the robot to manipulate two wafers simultaneously. The robot prefetches a wafer for a chamber before it removes the other wafer that occupies a chamber. This reduction in cycle time requires the robot controller to accurately track wafer presence or absence on each of the blades and maintain constant acceleration rates and velocities. This can save several seconds for each wafer exchange, increasing tool throughput. For example, if there is a savings of 2 seconds per exchange for a process tool with a current rate of 60 wafers/hr, the time savings is 120 seconds per hour. In this 2 minutes, two additional wafers can be processed each hour.⁹

Vacuum Processing Challenges

Lubricants are easily broken down by process chemicals. Lubricants must not cause outgassing or vaporize in a vacuum. Polymers used as friction materials in bearings or wafer pads must be able to withstand chemicals and elevated temperatures. Most robots attempt to isolate the motors from the vacuum environment. This requires mechanical motion to be transmitted from the atmosphere to the vacuum environment.

Ferrofluid seals can be used to pass rotary motion between the two environments. These seals must be protected from solvents used to clean or decontaminate the chamber. If the system is sealed at the vacuum chamber interface, motors and main support bearings can be placed outside the vacuum environment. An alternative is magnetic coupling. Magnets are used to transmit torque or forces from the atmosphere across a thin vacuum containment wall to shafts in the vacuum. Although magnetic coupling provides very good isolation, it also limits access. The physical vacuum containment wall prevents the routing of data or power lines to the end of the robot. Magnetic coupling can also reduce the stiffness of the system, making smooth control at high speeds difficult. Peak force, and therefore speed, is limited by the magnet's ability to transmit force across the air-gaps and vacuum wall. If too much force is applied, magnets slip and lose their position or act as springs. Furthermore, the shafts on the vacuum side of a magnetically coupled system must be supported by bearings. This can be an added source of particulate generation or bearing failure, reducing overall system reliability.

To build equipment for a vacuum environment is a rigorous engineering feat. Ideally, robot arms should feature few bearings, and they should be fully enclosed. Below the vacuum chamber, direct-drive capabilities contribute the optimum in controlled power. Smooth movement replaces the jerky motions of the past. Robots operating in vacuum create a variety of new problems and challenges for both mechanical and electrical structures. Mechanical structures are redesigned to eliminate small semienclosed areas that create leaks, while screws and blind holes are vented. All exposed metal surfaces, including electrical contacts and magnetic materials used in motors and solenoids, must be resistant to the effects of gas.

Because of the absence of air, the control system of a robot plays a critical role. Two approaches can be used to secure wafers for movement: passive and active. In the active case, an electromechanical assembly is used to grip the edge of the wafer. If too much grip is used, the wafer is distorted and, if the gripping fixture moves out of the exclusion zone on the wafer edge, yield loss results. A passive approach lifts the wafer and transports it using the coefficient of friction. This passive system requires the wafer to travel at speeds that prevent it from slipping.

Active gripping is an emerging method within vacuum, although atmospheric robots have used this method for several years. Active grippers use either mechanical or electrical mechanisms to constrain the wafer on the blade. The advantage over the passive method is that friction between the robot blade and wafer created during passive gripping prevents slippage, but it also limits maximum acceleration and speed. Experiments performed by Smart Machines (San Jose) have found the coefficient of friction between a wafer and a metallic blade to be as low as 0.24 Gs. This limits acceleration values to <0.24 Gs (2.34 m/sec²). The coefficient of friction can be increased by adding elastomeric pads. However, in many cases, pads cannot be used to increase the friction constant because of high temperatures or chemicals used in the process chambers. Active gripping can increase acceleration rates to more than 2 Gs on modern direct-drive robots, greatly reducing cycle times.

Conclusion

Both the mechanical and control architecture of vacuum robots are driven by customer requirements. As wafer sizes continue to grow, new processes evolve and handling procedures advance. Product lines must be capable of being modified to support both changing and emerging requirements. This flexibility adds substantial burden on the robotic control system. Not only must the system reliably position wafers, it must optimize trajectories, be scalable, support multiple wafer sizes, adapt to emerging communications standards, and behave gracefully, if not predictably, in critical situations caused by power failures and emergency stops.

Flat-panel display (FPD) and hard disk drive (HDD) equipment manufacturers have similar issues as semiconductor equipment manufacturers. Process tools need only be scaled to larger sizes for the FPD market and smaller sizes for the HDD market. Cleanliness, high reliability, and the way products are transported through the process is the same in all three cases. With open architectures, some companies are taking an open approach to the way they create robots that can embrace new standards, new sensors, and new media, while continuing to enhance—not replace—their basic architecture.

References

1. Hutcheson D, private teleconference, November 1997.

2. National Technology Roadmap for Semiconductors, San Jose, SIA, 1994.

3. Richard G, "Impact of the 300-mm Product Transition on the Semiconductor Industry," San Francisco, Hambrecht & Quist, August 11, 1997.

4. Van Zant P, "Microchip Fabrication: A Practical Guide to Semiconductor Processing," New York, McGraw-Hill, p19, 1990.

5. "Airborne Particulate Analysis for the Smart Machines SVR 200/300 Robot," Fremont, CA, Dryden Engineering, January 26, 1998.

6. Handbook of EU EMC Compliance, Boxborough, MA, Compliance Design, 1996.

7. Electrical Product Acceptance in Europe, Rosslyn, VA, National Electrical Manufacturers Association, 1995.

8. NFPA 79 Electrical Standard for Industrial Machinery, Batterymarch Park, MA, National Fire Protection Association, 1994.

9. Aalund M, "Design and Development of a Multi-Channel Robotic Controller," master's thesis, Department of Mechanical Engineering, Austin, TX, University of Texas, 1991.

Martin P. Aalund, PhD, is the engineering director at Smart Machines (San Jose). He is responsible for providing leadership in open systems architecture for robotic controls and developing state-of-the-art control technology for direct-drive vacuum robots. Before joining Smart Machines, Aalund was the chief technology officer at ARM Automation, a small, high-technology start-up that develops modular actuators for robots and distributed intelligent control systems. Aalund holds a BS, an MS, and a PhD in mechanical engineering from the University of Texas, Austin. He currently serves on the board of directors of ARM Automation. (Aalund can be reached at 408/324-1234, ext. 169.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit