Design of a surface inspection tool for rapid detection of micrometer-sized water droplets on a silicon wafer

Abstract

In this design study a new design is proposed for an in situ surface inspection tool that can rapidly detect micrometer sized water droplets on a silicon wafer. This surface inspection tool is part of a proposed redesign of the thermal conditioning stage of VDL-ETGs atmospheric wafer handler. A meticulous requirements study was performed, and the most important results were the following: 1) droplets with diameters larger than 35 μm should be measured, 2) the size, shape and location of the droplets is not relevant, 3) the wafer is not allowed to heat up by more than 3 mk as a result of the inspection, 4) the entire wafer surface should be inspected in less than 2 seconds and 5) only one droplet may pass undetected in every 6200 wafers. Several concepts were generated and after comparing them the synchronous dark-field flying spot scanner was selected, mainly for its high signal strength, and the low amount of signal fluctuation for different droplet locations. The design contains several parameters, the combination of which influence the system performance in a non-trivial way. A parametric design tool has been developed that takes in the design parameters as inputs, and outputs the system performance values and important design dimensions. This design tool was used to find design parameters that yield the desired performance. A simplified version of the design was built as an experimental setup for validation. For the experimental setup, a small spot size was chosen. This allowed us to better characterize the scattering behavior, however as a consequence this also prevented us from directly comparing the experimental results to the system performance yielding from the design tool. The experimental results were the following: 1) A very small part of the droplet surface is responsible for the large majority of the measured scattering, 2) the peaks are narrower in the cross-scan direction than in the scan direction, most likely due to the droplets being flat. 3) The measured noise is much higher than the predicted noise, this is likely due to poor grounding, poor shielding or a lack of filtering out high frequencies. Although the experimental setup yielded interesting results, the setup was not sufficient to fully validate the design – further experimentation is required to determine 1) if the predicted signal amplitude is correct, 2) if the signal amplitude scales with the square of the diameter of a droplet, 3) if the optical components remain as clean as we expect during operation of the dry unit, and 4) if the noise levels can be reduced to the desired level. Additionally, to improve the design further we suggest looking into ways of reducing the wafer warp and wobble.