Cover Article Equipment Frontiers
Robot for use in ultrahigh vacuum
Masafumi Kanetomo, Hideo Kashima, Central Research Laboratory, Hitachi Ltd., Tokyo, Japan
Takamichi Suzuki, Production Engineering Research Laboratory, Hitachi Ltd., Kanagawa, Japan
A new polar coordinate, wafer-transfer robot has been developed for thin-film manufacturing in an ultrahigh vacuum. The robot has a frictionless spring structure that operates in the vacuum and is magnetically driven by a motor located outside the vacuum. All moving parts in the vacuum environment have a minimum of two bearings. A repeat position precision of =0.2 mm in normal atmosphere (1.013 ? 105 Pa) as well as a maximum transfer distance of 800 mm and a transfer speed of 400 mm/sec, both at a pressure of 10-7 Pa, were reported. After operation in a cleanroom at atmospheric pressure, only three particles above 0.1 ?m in dia. in the revolving robot arm area were observed.
Figure 1. a) Wafer-transfer robot with arms stretched; b) wafer-transfer robot with arms retracted.
A typical system used to grow thin films epitaxially consists of a reaction chamber with a minimum pressure of 10-8 Pa and a wafer-transfer chamber with a pressure of 10-6 Pa [1-4]. Since the films are grown in the reaction chamber after the wafer is transferred from the transfer chamber, a transfer mechanism is thus necessary. Because this configuration tends to be rather large, several reaction chambers are now commonly used. The wafer-transfer system used should be able to operate in an ultrahigh vacuum, generate few particles, and occupy little space .
When a machine-driven system is used in a vacuum, lubrication of its ball bearings is critical to contamination-free operation. A vacuum grease with low vapor pressure can be used at pressures above 10-4 Pa as a lubricating material. An ultrahigh vacuum, however, cannot use vacuum grease because the chamber and the transfer system must be heated to 150-250?C to maintain the high vacuum. A solid lubricating material is thus more suitable because of the small amount of gasses it releases. Since the exact amount of solid lubricant used must be measured carefully, after calculating the load supported by the ball bearings, the total system is likely to become complicated if many ball bearings are used.
Structure of thewafer-transfer robot
We have developed a polar coordinate, center robot (Fig. 1) for a multichamber molecular beam epitaxy (MBE) system that has spring arms made of sheet springs and rigid plates. The arms operate in the reaction chamber and are magnetically coupled to a driving motor outside the vacuum area. Only two ball bearings are used in the vacuum area to rotate the arms; there are no ball bearings in the arms . A multichamber ultrahigh vacuum system using this wafer-transfer robot occupies 30% less square footage than a conventional linear-magnet-coupled, wafer-transfer robot.
Figure 2. Diagram of wafer-transfer robot.
Each spring arm is attached to a revolving rigid arm on the robot control structure (Fig. 2). This structure is attached to a conflat-type flange with a 300-mm dia., sealed with a copper gasket. The materials used for the parts in the vacuum area are either stainless steel or aluminum and can withstand baking at 250?C. The revolving rigid arms can be operated independently from outside the vacuum. A holder attached to the top of both arms positions the wafer. When the spring arms move in the same direction, the holder rotates around the central axis. When the arms move in opposite directions (one clockwise, the other counterclockwise), they change their shape and the holder moves toward the central axis. The spring arms move linearly via contractions and expansions in the radial direction. The robot revolves when the arms shrink.
Figure 3. Transfer of wafer using wafer-transfer robot. The sequence of transfer is illustrated in a)-f).
A pulse motor is used to drive each rigid arm. The revolutional power of each arm is transferred through a driving gear into an introduction shaft connected directly to the reduction gear. Each shaft is connected to both a driving ring, supported by a ball bearing, and a magnetic coupling that transfers the revolving movement into the vacuum region. The magnetic couplings use a permanent magnet outside the vacuum to transfer the rotation to a permanent magnet inside the vacuum. When the exterior magnet moves, the interior magnet moves. With this structure, the two pulse motors can independently operate the two rigid arms.
This wafer-transfer robot moves a wafer in a conventional MBE vacuum system composed of three chambers (Fig. 3). Chamber A is the transfer chamber and chamber B is a reaction chamber. The transfer chamber, between chambers A and B, contains the robot. Transfer occurs as follows:
a) The wafer holder turns toward chamber A by the revolving motion of the arms.
b) The holder moves into chamber A by the linear forward movement of the arms. The holder slides under the wafer on the table.
c) The holder with the wafer moves out of chamber A by the linear retraction of the arms.
d) The holder with the wafer turns toward chamber B by the revolving motion of the arms.
e) The holder with the wafer moves into chamber B by the linear forward movement of the arms. The wafer on the holder is moved onto the table.
f) The holder is moved out of chamber B by the linear retraction of the arms.
With this series of repetitive movements, a wafer can be easily moved into any chamber adjoining the transfer chamber. Figure 1b is a photograph of the robot with its arms retracted.
Structure of the spring arms
Each spring arm consists of two wavelike sheet springs made of stainless steel, one on the upper and one on the lower side. Parallel sheet springs (Fig. 2) are connected to the end of each spring arm to support the wafer holder. Each arm has the following features:
It maintains strong horizontal rigidity even when fully extended.
The revolutional load torque generated by the reaction force of the spring arms is lower than the magnetic coupling torque.
The maximum stress generated by changing the shape of the sheet springs is lower than the maximum permissible stress.
The movement of the holder supported by the two spring arms is smooth and linear.
The arm position in the radial direction is highly reproducible.
It operates cleanly, i.e., particles are not generated by the robot movement and changes in configuration.
The wavelike shape of the spring arms strengthens the horizontal rigidity when the arms move linearly in the radial direction. The shape of the sheet springs was optimized by simulating a large deflection model of 3-D shell elements and using the finite element method.
Since the arms are symmetrical, it was only necessary to analyze the change in shape of one. We simulated the relationship between the displacement of the arms and the linear translation of the holder. We also calculated the stress generated in the sheet springs when they changed their shape and the horizontal rigidity at the holder. Figure 4 shows the calculated linear movement of the holder as a function of the rotational angle of the rigid arms. As the rotational angle increases, linear movement along the radius increases proportionally. However, with a large rotational angle, linear translation displacement is reduced.
Figure 4. Calculation of wafer-transfer-robot motion.
Figure 5. Construction of magnetic coupling.
Structure of the magnetic coupling
Figure 5 shows the construction of the magnetic coupling that transfers the driving force from outside the vacuum into the vacuum. As the exterior magnets move, the interior magnets also move. When load torque is generated at the interior magnets by the reaction force of the spring arms, they lag behind the exterior magnets. The coupling torque must be larger than the load torque. The coupling torque transferred from the motors depends on the positional lag and its strength is determined by the magnetic flux density, which depends on the gap between the magnets. A higher magnetic flux density generates a larger torque. The torque can also be increased by using larger permanent magnets, but this requires more space. We therefore used Sm-Co magnets that contain a rare earth element, and which generate large magnetic energy despite their small size. Figure 6 shows the calculated relationship between the lag angle for the paired magnets and the coupling torque. A transfer torque of 100 kg-cm was generated at a relative lag angle of 5? while the maximum transfer torque was 160 kg-cm at 15?.
To evaluate the performance of our robot, we measured the reproducibility of repetitive rotation sequences performed by the robot in atmosphere. The revolution position of the holder was determined at each transfer sequence point by a laser measuring the displacement. The resolution power was 5 ?m. Figure 7 shows the reproducibility was within 0.2 mm for over 125,000 trials.
Figure 6. Calculated relationship between lag angle for paired magnets and coupling torque.
Figure 7. Reproducibility of repetitive rotation sequences.
When the spring arms are retracted, the distance from the center of the revolution to the edge of a 200-mm wafer on the holder is 430 mm, so the diameter of the transfer chamber housing the robot needs to be at least 900 mm.
To evaluate operational cleanliness, we installed the robot and a laser dust monitor in a Class 1 cleanroom (Fig. 8a). The suction-type, laser dust monitor has a flow rate of 2.8 liters/min. The 10-mm-dia. hose for sampling was placed under the revolving arm of the robot and particle generation was measured during each transfer sequence. Only three particles larger than 0.1 ?m in dia. were found during 25 hr of measurement (Fig. 8b).
To determine the extent of pressure reduction, we set the robot in an ultrahigh vacuum chamber and pumped the air out. The system was then baked at 250?C for 24 hr to reach an ultrahigh vacuum. This was followed by more pumping for 48 hr, which reduced the pressure to 1.0 ? 10-7 Pa.
Figure 8. a) Setup for measuring operational cleanliness and b) results.
We measured the residual vibration of the holder after revolution. The initial amplitude of the vibration was about 0.2 mm, which dropped to about half after approximately 10 sec (Fig. 9a). The vibrational frequencies were about 5.5 Hz. Figure 9b shows the residual vibration at the holder position after the spring arms were stretched out by linear forward movement. The initial amplitude of vibration was 0.02 mm, which dropped to 0 after approximately 0.2 sec. The vibration frequencies were about 75 Hz.
The robot arm can be considered as a vibration model consisting of a mass and spring. Vibration results from the rotational stiffness of the magnetic coupling and the weight of the moving system (rigid arms and spring arms). The higher the spring stiffness, the smaller the vibration amplitude. High stiffness of the magnetic coupling is needed to reduce the vibration amplitude and to accelerate the attenuation of the vibration. To obtain this stiffness, the gap between the magnets should thus be reduced, and magnetic materials having a high BmaxHmax (max. magnetic flux ? max. magnetic field strength) should be used.
Earlier vibration reduction results in faster wafer movement. To reduce the vibration time, the first order characteristic (resonance) frequency should be increased. The weight of the total system should be reduced and its stiffness should be increased. For our robot, the characteristic frequency for revolution was small when the arms were shrinking, while, in comparison, the characteristic frequency in the radial direction was quite large. Therefore, the arms can move faster in the radial direction than in the direction of the revolution.
Next, we measured the coupling torque and relative lag angle of the magnets. When the lag angle was small, the stiffness, defined as the coupling torque divided by the angle of displacement, was 21 kg-cm/? and almost linear. The maximum load torque of the spring arms was 52.4 kg-cm at a relative lag angle of 2.2?.
Figure 9. Measured residual vibration in holder a) after holder revolution and b) after linear forward movement of arms.
Since this robot is still in the prototype stage, its cost is unclear. However, to minimize the production cost, the number of parts should be greatly decreased (from 40) either by making the sheets as one component or reducing the number of parts in the arms.
Application to thin-film manufacturing devices
Figure 10a shows the overall plan for application of our wafer-transfer robot to a thin-film manufacturing device in an ultrahigh vacuum. The transfer chamber including the wafer-transfer robot is attached to the reaction chamber through a gate valve. The stroke of the robot is 700 mm, which corresponds to the distance between the reaction chamber and the transfer chamber. Figure 10b shows the overall plan for a conventional wafer-transfer system with a linear magnetic coupling. Almost 1000 mm of extra space is needed for the coupling, contributing to the large space needed for the total system. Our proposed system reduces the required floor space by about 30% compared with the conventional system.
Figure 10. a) Overall plan for wafer-transfer robot applied to a thin-film manufacturing device in an ultrahigh vacuum, and b) overall plan for conventional wafer-transfer system with linear-type magnetic coupling.
We have developed a polar coordinate, wafer-transfer robot that uses spring arms made of sheet springs and a revolutionary
magnetic coupling system. It has a repetitive position precision of =0.2 mm in atmosphere, a maximum transfer distance of 800 mm, and a transfer speed of 400 mm/sec in vacuum (10-7 Pa). After 25 hr of operation in a cleanroom at atmospheric pressure, only three particles larger than 0.1 ?m in dia. were detected. This transfer robot is applicable to the manufacture of thin-film devices in ultrahigh vacuum.n
We are grateful to Tetuo Ito for his kind assistance with the computer simulation used throughout this work.
1. J. Sakai et al., "High-throughput and Fully Automated System for Molecular-beam Epitaxy," Journal of Vacuum Science and Tech., Vol. B6, p. 57, 1988.
2. T. Sonoda et al., "Ultrahigh Throughput of GaAs and (AlGa)As Layers Grown by MBE with a Specially Designed MBE System," Japanese Journal of Applied Physics, Vol. 27, p. 337, 1988.
3. K. Kondo et al., "MBE as a Production Technology for HEMT LSIs," Journal of Crystal Growth, Vol. 95, p. 309, 1989.
4. T. Sonoda et al., "Ultrahigh Throughput of GaAs and (AlGa)As Layers Grown by MBE with a Specially Designed MBE System," Journal of Crystal Growth, Vol. 95, p. 317, 1989.
5. M. Kanetomo, H. Kashima, T. Suzuki, "Wafer-Transfer Robot for Use in Ultrahigh Vacuum," Journal of Vacuum Science and Technology A, Vol. 15, No. 3, Part 2, pp. 1384-1387, 1997.
6. "Shudobuganai Robot Arm" ["Frictionless Robot Arm"], Nikkei Mechanical, p. 79, May 17, 1993.
For more information, contact: Masafumi Kanetomo, Central Research Laboratory, Hitachi Ltd., 1-280 Higashi-koigakubo, Kokubunji-shi, Tokyo 185, Japan; ph 81/423-23-1111, ext. 2120, fax 81/423-27-7727, e-mail Masafumi-crl.Kanetomo@c-net3.crl.hitachi.co.jp.