Automated Inspection of Quartz Crucibles - For use in the Czochralski Process
Abstract
As the energy consumption in the world continues to increase it becomes more important to find alternative sources of electric energy. Photovoltaics is a method which utilizes the sun to create electricity. The Czochralski process is used to produce high efficient mono-crystalline silicon ingots, which when cut into wafers are used in solar panels. A silicon seed is dipped in a quartz crucible containing silicon melt and pulled up, resulting in an ingot. An important parameter regarding the quality of the mono-crystalline silicon is the quartz crucibles used in the Czochralski process and it would be useful to be able to inspect a crucible before it is used. The aim of this thesis is to create a prototype inspection robot which can complete these inspections automatically.
A quartz crucible consists of two layers. Although the two layers consist of the same material, their properties are different. The inner layer is smooth and solid while the outer layer is rough and contains bubbles.
Two important parameters of a crucible which the inspection system developed in this thesis explores are impurities in the quartz crucibles and the geometry of the crucibles. Crucibles often has impurities represented by visible black spots. As crucibles deteriorate during the Czochralski process it is important to have as little impurities as possible to achieve a pure silicon ingot. Bad geometry properties in the crucible may cause bubbles or turbulence in the silicon melt. This can cause structural losses in the crystal making the ingot unusable.
The black spots are detected by a camera. The camera takes pictures covering the whole inside surface of the crucible. An edge detection function is used to separate the black spots from the rest of the crucible in the images. External lighting is used to clearly separate the black spots from the much lighter quartz. The inspection system detects the correct number of black spots with an accuracy of 95 percent. This is satisfactory and it is difficult to achieve a better result. This is satisfactory because some bubbles are very similar to the smallest black spots in an image, making it difficult to separate the two.
A confocal chromatic sensor is used to collect samples of geometric data. The sensor is able to measure the thickness of clear glass objects, making it a suitable choice for measuring the thickness of the bubble free layer. It can also measure distances to a surface. It is therefore suitable to measure the thickness of both layers and the diameter of the crucible.
The thickness measurement of the bubble free layer is fairly accurate and clearly shows a trend in thickness along the profile of the crucibles. The absolute thickness measurements are not entirely correct as the refraction index of the material has to be derived to achieve this.
The total thickness, thickness of the bubble layer and the diameter are not as accurately measured as the thickness of the bubble free layer. This is probably the result of the rough surface of the outside of the crucible and inaccurate calibration of constants used in calculating these parameters.
A robotic arm with six degrees of freedom is used to carry out the inspection process. Inaccuracies in the robotic arm give roots for inaccuracies in the system. Restrictions in the arm makes the system slower than what would be optimal and limits the area of the crucible which can be inspected. As a conclusion a totally new system setup is suggested, discarding the robot and utilizing a more custom design. This system would have the benefit of being both faster and more accurate.