The traditional Atomic Force Microscope (AFM) is a two-dimensional tool, which can only generate surface profile with limited depth variation. A design modification to the traditional AFM has been proposed by Jayanth et.al [1] to make it a true three-dimensional tool. The modified cantilever has a magnetic particle attached to it. These particles are attached indigenously on the commercially available cantilevers. The magnetic particle attachment process is very sensitive to vibration and requires very precise motion control avoiding any unnecessary body movement. Since conventionally it has been done manually, it imposes stringent constraints on the user performing this task. A more accurate and faster process was required to replace the existing system. This thesis develops an automated process for attaching a magnetic particle to the AFM cantilever. The new process requires very little manual involvement in the most critical steps of the process.
A setup was designed to incorporate the visual feedback from a camera attached to a microscope and actuation using a three-axis piezo stage. The image formation process was modeled relating any three-dimensional point to its corresponding location in the image captured by a Charge Coupled Device (CCD) chip. Image processing algorithms were developed to locate the particle, micro-pipette and the AFM cantilever, also to track the particle and micro-pipette in real-time. Finally a control system was designed which would control the location of the object (micro-pipette or the particle) by actuating the piezo stage. The control system used image from only one camera, while the hysteresis in the piezo stage, which was being operated in open loop, was countered for.
The control system was tested for automating different steps in the process. It was first calibrated to extract the necessary system parameters. Its robustness was tested by performing the process under varying illumination and vibration. The process could successfully attach particles of size ranging from 25 – 70 µm under these conditions. The automated process was twice as fast as the manual process and required minimal manual involvement in the particle attachment step. Similar results were observed for other steps that were automated, specifically the particle pick-up and glue dabbing step.