The expanding role of robots in process tools productivity
01/01/1999
ROBOTICS AND AUTOMATION
The expanding role of robots in process tool productivity
K. Charles Janac, Smart Machines, San Jose, Calfornia
While the results of wafer processing are the most important criteria in a production process tool, productivity is clearly the second. Process tool automation - the suite of critical wafer-handling components (i.e., atmospheric and vacuum robots, elevators, loadlocks, and pod openers) that move wafers inside and out of wafer-processing tools - is a key ingredient of tool productivity.
Automation can account for up to 33% of the parts cost of a wafer-processing tool; these components can have a major impact on reliability and productivity. In addition, the importance of process tool automation is growing because:
more tools are automatically rather than manually loaded from atmospheric buffer stations;
more processes, even lithography and metrology, are done in vacuum environments requiring robotic handling;
heavier payloads associated with 300-mm wafers drive the use of intrabay automation; and
cleanliness requirements continue to drive people out of wafer fabrication facilities.
Based on the above trends, spending for process tool automation is forecast to increase from today`s ~$34 million to $60-70 million/fab within the next four years.
The productivity elements of process tool automation (i.e., reliability, performance, ease of use, and cost) have been continuously improving; we can see this by comparing today`s robotics for wafer processing with those of just three years ago (see table).
Process tool automation has to be cost effective, very reliable, high performance, flexible, and ultraclean. Some of its key requirements become more stringent with each successive generation of process tool equipment.
 |
10 million MCBF reliability
From the broad list of productivity benefits, some are key. For example, as more processes are clustered in tools and intrabay automation, robotics has to be taken out of the reliability equation.
Consider that a typical etcher in a fab will execute ~1 million pick-or-place cycles/year to process 125,000 wafers. If the central robot has just a 1-million mean cycles between failure (MCBF) reliability, it will contribute one downing event from robot failure/year. If this particular etcher has an atmospheric load-unload robot and a vacuum robot of similar reliability, it will contribute two downing events/year. While this may not seem too frequent for a single tool, consider that in an average production fab, there are 170 vacuum robots and 542 atmospheric robots. This adds up to 712 downing events/year due to robot failure alone. (This example does not even consider elevators or aligners.) If it takes eight hours to recover from a vacuum robot failure and four hours to recover from an atmospheric robot failure, the 1-million MCBF has escalated to 1360 hr of lost production/year due to vacuum robot failures, and 2168 hr due to atmospheric robot failures. At $3000/hr, the total is $10 million in lost production/fab/year.
Now, run through this brief scenario for an ion implanter that executes over 4 million pick-or-place cycles/year. The obvious conclusion is that a 10 million MCBF is the better reliability target for process tool automation robots. This can only be achieved through testing the critical components, such as wire harnesses, amplifiers, ferrofluid seals, and bearings to several times the required reliability of the overall system. Consistent performance is also important. With today`s technology, robots can be designed to avoid continuous maintenance to keep a robot "in spec." In the past, for example, belt and gear wear changed performance over time, but today`s direct-drive robots using noncontact motors do not wear as much, thus helping to maintain consistent performance over the life of the robot.
Wafer exchange in productivity equations
As clustered process chambers increase and process times decrease, process tool automation becomes an increasingly important part of the throughput equation. Wafer-handling throughput is driven by scheduling through the tool, by the time robots need to move wafers, by the number of robot moves necessary to execute a wafer exchange (obviously fewer is better), and, ultimately, by robot architecture and the acceleration of each wafer move.
When considering wafer-handling architecture in process tool automation, there are several trade-offs. A single-blade robot has small swing clearance, but requires 15 moves for a simple wafer exchange A dual-blade robot (i.e., one that moves two wafers at once) requires 11 wafer moves, but has a larger swing clearance, making the process tool larger. A dual-powered, end-effector robot has the smallest number of wafer moves at 6, and the lowest swing clearance, balancing the trade-offs to optimize wafer throughput.
Consider, also, robot speed and its effect on throughput: Vacuum robots handling wafers at high temperature are only able to move at 0.15 g due to friction contact via low coefficient of friction, end-effector materials able to withstand 400?C wafer temperatures. Atmospheric robots with vacuum- or edge-grip end effectors are able to move at up to 1.5 g, 10 times faster. Robots with continuously optimized acceleration trajectories, one of the advances provided with today`s technology (see table), enable faster movement at any commanded acceleration than robots with trapezoidal accelerations able to ramp speed in linear segments. Continuously optimized acceleration trajectories require sophisticated algorithms, as well as powerful digital signal processors for trajectory computations.
Minimizing robot moves through innovative robot architectures, as well as accelerating at their allowed potential, improves the productivity of process tool automation-limited process tools.
As the table above illustrates, there are several other factors in the productivity benefits of process tool automation:
Robot repeatability. The ability to deliver wafers to a process tool chuck or loadlock within ?0.002 in. is crucial because greater variability may result in problems, such as the coating of edges of electrostatic process tool chucks, slowing down metrology stages, and touching the sides of cassettes. Most robots initially meet these repeatability requirements, but may degrade over time. Variation of repeatability can result in significant process tool downtime as service technicians diagnose and repair repeatability excursions. Direct-drive robots eliminate this variability, since their initial performance will be maintained during their useful life.
Vibration. End-effector vibration can cause backside wafer damage and particle generation. In the vacuum environment where friction contact is the means for holding the wafer, excessive vibration slows a robot to prevent the wafer from "walking." In general, vibration should be kept under 1:3 in the z-direction of the vibration during an extension; if maximum command acceleration is 0.15 g, vibration peaks during an extension move should not exceed 0.15 g. During such a move, vibration in the z-direction should not exceed 0.05 g. This ratio should be maintained for faster robots, such as atmospheric load-and-unload robots.
Microcontamination. Rotary bearings are the most crucial component in determining low particle generation of a robot. The constant motion of these metal balls against inner and outer rings and spacers generates particles. Bearings must be engineered for minimum particle generation and for minimum numbers in robot arms; further, they must be encased deep within the arm mechanism. Mechanisms, such as linear ball screws in the arm to achieve radial motion, must be avoided. Technical tricks, such as low air pressure inside the arm, can also be used to keep particles away from wafers. Another source of particles is motor brushes. While it is possible to seal the motors, particles eventually find a way out, creating a particle load problem for the fab filtration system. Brushless motors avoid this problem, and eliminate the need to replace motor brushes during the life of the robot as well.
Setup and error recovery. The faster robotics can be brought on line, the faster the process tool can become operational. This is true for both the initial setup and after recovery from an error condition. User-friendly software, high-productivity teaching schemes, and automatic error recovery can mean the difference between satisfied and frustrated users. One of the functions of process tool automation is to recover from faults that occur during the manufacturing process. Such faults include power losses, emergency power shut downs, collisions with fixed parts of process tools during robot setup, and failure to meet preventive maintenance schedules. Key to proper error recovery is a robust mechanical architecture, protected electronics, and user- friendly software. Direct-drive robots with relatively few moving parts have a good chance of surviving collisions with chamber walls or gate valves. Absolute positioning capability allows robots to be aware of their position at all times and to restart without homing. Active braking schemes bring a robot to an emergency stop without losing the wafer.
Flexibility. There are more than 200 process tools in a typical production fab. The requirements of these process tools drive the architecture and performance of the associated automation. A variety of arm reaches, vertical travel heights, communication options, end-effector shapes and materials, and throughput requirements need to be met to achieve a solution for a variety of processes. Some process tools have cassette stations and process chambers oriented in a radial pattern, while others have them configured in a linear orientation. Other tools require wafers to be flipped during processing. Vacuum tools have different levels of vacuum from barely subatmospheric to 1 ? 10?9 torr. CMP and copper plating tools require robotics that are able to operate in a variety of liquid environments and are still able to function. The requirements are so varied that a library of mechanisms is required to fulfill them. The current efforts of SEMATECH and I300I to standardize automation for 300-mm applications will be a great help in reducing the large number of variations required to serve the current generation of process tools.
K. CHARLES JANAC is president and CEO of Smart Machines, 651 River Oaks Pkwy., San Jose, CA 95134; ph 408/324-1234, fax 408/324-1966,
e-mail [email protected].