Development of Micro-manipulation System for Operation in Scanning Electron Microscope H. Eda, L. Zhou, Y. Yamamoto, T. Ishikawa, T. Kawakami and J. Shimizu System Engineering Department, Ibaraki University, Hitachi-shi 4-12-1, Japan, 316-8511 Tel: +81-294-38-5188 Fax: +81-294-38-5188 Abstract The ultimate goal of this project is to develop a manipulation system enabling unskilled operators to deal with objects of micron or sub-micron size in a scanning electron microscope, as easily as to deal with objects in usual size. Described in this paper is the results achieved in the first phase of the research, in which the focusing point is given to the conceptual design, the prototype development and the operability evaluation. The system is modularized into the manipulation unit, the control unit and the man-machine interface. The manipulation unit is further comprised of a twinarm manipulator mounted on a rotary table and a specimen stage with four degrees of freedom linear along X, Y and Z direction, and rotational around the Z-axis. The manipulator is driven by PZT actuators with magnifier elements and able to cover an envelope as wide as 200µm for each axis of X, Y and Z. Instead of doing a direct operation, the operator steers the manipulator via an user-friendly interface which is designed to absorb the optical and mechanical variations. It allows the operator to concentrate to the manipulation without paying mach attentions to the changes in magnification of SEM or other conditions. The control unit merges the visual information of the SEM and the manipulation information from the user interface and derives the optimum locomotion of the arm for the desired operation. Keywords: micro-manipulator, micromachining, PZT actuator, scanning electron microscope, man-machine interface 1 Introduction In the applications of precision, medical and biological engineering, there are increasing demands of implementing fabrication, assemble, inspection, modification and/or evaluation at the micron or sub-micron scale 1) 2). Currently, most of such operations are manually done by highly skilled operators with the assistance of scanning electron microscopes (SEM). It is obviously subject to the magnification of the SEM. The limited visible area at a high magnification of SEM makes it difficult for the operator to find and trace the target. The operators have also to experience a long time training to familize the feeling of micro-scale movements in order to achieve high repeatability and high accuracy. Yet, the operations are tedious and time consuming. It is, therefore, strongly expected to develop a manipulation system possible for unskilled operators to easily execute the desired tasks. As the first step of research, a manipulation system (including hardware and software) has been designed and developed to smoothly and continuously execute a series of micro-operations such as cutting-grasping- transferring, at different SEM Control module Manipulator Mechanism module Vision system Mechanism module Fig.1 System concept Fusion Controller module Man-machine interface module Operation panel Man-machine interface module Fig.2 Main modules and key components 95
SEM magnifications. The scope of this paper included the conceptual design, the prototype development, field tests and the operability evaluation. Especially, a parallel plate type force sensor has been newly developed and incorporated into the manipulator arms to deal with the micro objects from as soft as tissues to as hard as metals, which has been considered difficult to realize with the conventional tools. 2 Conceptual design The system has been developed with a basic concept that the unskilled operators can intuitively work on the desired tasks without paying much attention to the changes in working conditions like the object size, SEM magnifications and etc.. Instead of manipulating the object directly, the developed system allowed the operator to have a good control of the manipulators via an interface with a great workability. The interface was designed to absorb the mechanical and optical variations by using the visual information and sense the tiny force exerting on the objects, to enhance the repeatability and accuracy. With such configuration, the developed system made it possible for the operator to concentrate on the tasks based on the visual information. converted into the resistance of the joystick so that the operator could exactly feel the interference between the manipulator and target. Afterward, the operation was carried out at the continuous visual and force feedback. 3 Module description 3.1 Mechanism module The detailed drawing of the mechanism module was schematically shown in Fig.3. The manipulator had a twin-arm configured in the leftright symmetry. Each arm came with three actuators superimposed one on the top of another in order to move along the XYZ axial direction. As shown in Fig.4, the actuator was driven by the piezoelectric element and its displacement was enlarged 20 times by a parallel plate structured magnifier. The working range was from 0.1µm to 200µm for each direction of X, Y and Z axis. Similar to the human hands, the manipulator took The conceptual layout of the system was shown in Fig.1 and Fig.2. The system was comprised of three main modules; the manipulation mechanism, the controller and the man-machine interface. Each module was responsible for different function. The mechanism module was set inside the vacuum chamber of the SEM for the object manipulation as well as the direct observation. The interface module with a great workability provided the visual information to and retrieved the instructions from the operator. The other tasks including the measurement, image processing and necessary force feedback were automatically executed by the control module. The mechanism module further incorporated a twin-arm manipulator, a movable specimen stage and a rotary table. Through the interface, the operator was able to remotely move the left and right arms as if his own. To begin with, the vision system of the control module captured the image via a the microscope (SEM), abstracted and displayed the information necessary for the feedback control and manipulation. Based on the visual information, subsequently, the operator gave instruction to the control module. By fusing the visual information and operational information that the operator gave, the control module derived the commanding information for the movements of the manipulators. At the moment the manipulator interacted with the target, the force detected was Twin arm manipulator Rotary table Z X Y Specimen stage Fig. 3 Manipulation mechanism PZT actuator PZT Fig.4 PZT actuator and magnifier twin-arm configuration. Each arm was equipped Instant rotational center DC Motor Roller pair Fig.5 Rotary table with the mechanism of rotating around any specific center 108
with a parallel plate type force sensor to detect the force ranging from 90µN to 30mN, which is response to the fabrication of soft/low strength objects and hard/high strength objects as well. The individual arm could play different role or cooperate each other for complicated tasks. For example, one arm held the object, while another arm executed the concrete operations such as cutting, peeling, removal and carrying. The manipulator was mounted on a rotary table. As illustrated in Fig.5, the table was supported by a pair of rollers set perpendicularly and driven by DC motors. The attitude of the roller pair was adjustable within 3 degrees by the PZT actuators together with parallel plate magnifiers. The point where the lines perpendicular to the roller pair intersect defined the instant rotational center of the table. It meant that the rotational center is controllable by the attitude of the roller pair, so that the table rotated around any specified center. With assistance of such a rotary table, the system was able to position the manipulator against the target from any desired direction. The specimen was seated on the specimen stage which is movable linearly at the X, Y and Z direction, and rotationally around the Z-axis. The Y DC motor for Z-axis Z DC motor for θ-axis X Fig.6 Specimen stage DC motor for Y-axis consideration resulted in a concept of that the operator, instead of doing a direct operation, handled manipulator via the user-friendly interface with great workability. Shown in Fig.8 was the operation panel of the interface module. Joysticks were chosen as the operation tool for its high performance. Out of four joysticks, two in most right and left were used for the control of the right and left arms. Their transverse and longitudinal travels were respectively synchronized to the movements of arms in X- and Y-axial direction. Fig.7 Manipulator assembled on SEM chamber kit The arms moved in the Z axial direction when the joysticks were twisted. The moving speed was subject to the magnification of the SEM, so that the manipulator always traveled with a constant speed cross the image display, regardless of the change in SEM magnifications. The movements of the rotary table and the specimen stage were automatically run by the control module, to visually captured the object, to locate it within the scope and to absorb various mechanical and optical variations. In case of manual operations required, the other two joysticks were available to provide the functions as same as mentioned above. This arrangement enabled the operator to intuitively operate the manipulator with feeling as same as he moves his own arms. Unlike the conventional one-way command input tools, the newly developed joysticks could movable range was ±3mm for each axis and 360 for rotation. The mechanism of the specimen stage was detailed in Fig.6. Fig.7 showed the manipulator module assembled on the SEM chamber kit. 3.2 Interface module Left arm controller Potential meters Contact indicators Right arm controller As described previously, the system development was emphasized on the point of how to realize the operations desired by the unskilled operators. This Rotary table controller Specimen stage controller Fig.8 Man-machine interface 109
bidirectionally communicate with the manipulation module 3) 4). It meant that the joysticks were able to give the instructions for movements while retrieving information from the manipulator. To remotely manipulate a tiny object in SEM, for example, it was effective to utilize not only the visual and position information but also the information of force exerted on the object. In this system, the strength of the working force was continuously measured by the strain gauges attached in the manipulator tip. By feeding the detected forces back to the joystick via a torque motor which could generate a constant torque proportional to the driving current, the user actually sensed whether the contact was made to and how much the force was applied. By feeding the detected forces back to the PZT actuator, additionally, the active force control has been realized to maintain a constant force between the manipulator and the object being manipulated. With force control, the contact point between the manipulator and object became floating in nature 5) so as not to undesirably damage the object, either to remove excessive material from the object. 3.3 Control module One major function of the control module was to automatically track and locate an arbitrarily shaped object when the visible area or the SEM magnification is dynamically changed during the operation. The current research has developed an image-processing algorithm to capture and track any specified object. The target, it could be on the specimen and/or the manipulator tip, was first defined or selected by simply clicking the mouse button and related to the actuator (the specimen stage or the rotary table) where it sat on. The developed algorithm then digitized the image captured and converted the pixel into the position information. Subsequently, when the SEM magnification changes, the control module drove the specimen stage or the rotary table to keep target(s) always within the scope. In addition to the auto-track function based on the visual information, the control module also interpreted the instructions given to the manipulator by the operator into the appropriate movement and speed to match the current SEM magnification. It offered a flexible environment for the operator to work easily yet efficiently. Shown in Fig.9 was the information flow in the control module. The high performances carried out in the controller were all realized on a personal computer. Such environment made it possible for operator to intuitively work out the tasks based on the visual and force information, even without feeling changes of the magnification. 4 Field tests In order to evaluate the capability of the developed system, several functional tests were carried to perform cutting, marking, transferring, moving, lifting and flipping-over. Two of them, cutting and peeling/transferring, were demonstrated below. 4.1 Cutting The object shown in Fig.10 (a) was a star-shape cell of goumi leaf. The manipulators were equipped with sharp needles as their operation tools. The test was designed to investigate the cooperativity of the twin-arm manipulator and specimen stage. The required task was to cut the cell into half, which is one of the operations most frequently performed to study the physical/chemical properties inside cells. SEM Image processing Visual information Monitor Twin arm Manipulat or XYZ XYZ Strain gauge PZT Strain amp Actuator driver A/D D/A Amp Torque motor XYZ XYZ Rotary table ω θ DC Potenti motor al meter PZT Amp Actuator driver D/A CPU A/D XYθ Joy stick Specimen stage XYZθ DC Potenti motor al meter Amp Motor driver XYZθ Fig.9 Information flow 110
Epidermis Star-shape cell Left arm (a) Tool tip and target Right arm 100µm (a) Contact 100µm (b) Approaching while visual area changed (b) Stripping Cell cut-off (c) Right arm holds while left arm cuts the target (c) Lifting (d) Target cut off Fig.10 Field test 1: cutting (d) Transfering Fig.11 Field test 2: Peeling/transferring 111
In the photo (b), the target cell was manually moved to the center of the scope and the manipulators were positioned at where the operation was intent to start with. Once the target and the relevant tool tip were specified and selected, they were always located within the SEM scope by the auto-track function. Regardless the changes in the SEM magnification, the operator felt no visual lag and could immediately start the necessary operation. In the photo (c), the right arm came to hold the top half of target cell, while the left arm cut off the other half. The photo (d) showed the cell after completion of the cutting. 4.2 Peeling/transferring The objective of the experiment shown in Fig.11 was to latch the epidermis of the botanical tissue, strip it off the body and transfer it to a required place. The task placed the importance on cooperation between arms and force feedback. As shown in (a), the specimen stage was first shifted to locate the point for operation. Then, the left arm was made a contact to the specimen and latched the targeted substance. Occasionally, this operation was possibly completed by the left arm only. For most cases, however, it required cooperative operation between the manipulators, rotary table and specimen stage to properly position both manipulators arms and specimen. In the photo (b), the right arm came to help the left arm to hold the target. The subsequent striping task was intuitively done by lifting both arms together. It was very difficult for conventional manipulators to keep the relative position of tips unchanged while moving arms in 3-D space. For this case shown in (c), the active force control was utilized to sustain the object, with the newly developed algorithm which was able to establish a constant force for the interaction between right and left arms. After being separated from its body, the target epidermis was transferred to another place. During the transferring, as shown in (d), the relative position between the right arm and the left arm was frozen by the active force control. Also, the autotrack function made the target always visible when the arms as well as the target traveled a long distance even exceeding the current scope of the SEM 5 Conclusion This paper has described the micro-manipulation system developed for operations in SEM environment and its performance in the field test. The developed system united both observation and manipulation together, which are normally done in different platform, in a single set-up. It allowed the operator to perform necessary fabrications during the observation, also to monitor the manipulation on-site. The achievements were summarized as follows; (1) A twin-arm micromanipulation system, which was able to cooperatively manipulate the objects from as hard as tissue to as soft as metals, has been developed for SEM environment. (2) A parallel plate type force has been developed and cooperated into the tool tip to sense the interfering force between the tool and specimen. The force feedback control not only provided the actual resistance to the maninterface (joystick), but also actively controlled the force being constant during the operation. (3) The image processing algorithm has been developed to countermeasure the missing target problem. The vision feedback system could automatically track any specified target and locate it with in the scope of SEM. The instructions governing the manipulator movement were also translated to an appropriate speed and distance based on the visual information, to match the current SEM magnification. The field tests had shown that the basic concept of the developed system is applicable to the micromanipulation such as cutting, grasping, peeling and transferring. The future research will include skill analysis, knowledge base establishment, auto/manual mode development and man-machine interface evaluation 6). Acknowledge This work was financially supported by the Ministry of Education, Science and Culture (Japan) under the Fundamental Research Type A (10305012). References 1. Marc Madou: Fundamentals of Micro Fabrication, CRC press, 1997 2. S T Smith and D G Gatwynd: Fundamentals of Ultra Precision Machine Design, Gordon and Breach Science Publishers, 1992 3. Dieter Vischer: Cooperating Robot with Visual and Tactile Skills, Proceedings of IEEE International Conference on Robot and Automation, 1992 4. S Hirai, et. al: Integration of Task Knowledge Base and a Cooperative Maneuvering System, IEEE International Workshop in Intelligent Robots and System, 1990 112
5. T L Graf: Deburring, Finishing and Grinding Using Robots and Fixed Automation: Methods and Applications, Technical paper, 3M Abrasive System Division,1993 6. H Fujita: Micro Machine World, Kogyo Chousakai,1993 113
114
95 95