Factors affecting probability distribution and yield of silicon dioxide defects

It is now possible, using nondestructive liquid crystal technique, to test the distribution and the yield of a particular but important type of defect across the wafer surface, the oxide defect. It is shown that, for defects in silicon dioxide grown in dry oxygen, the probability distribution obeys Poisson statistics for different chip sizes, different times of field application, different oxide thickness and different values of applied field. The average number of defects per chip increases exponentially with the increase in the fourth root of time of field application until saturation. The average saturation number of defects per chip exponentially decreases with the increase in the square root of oxide thickness. For both cases - before and after saturation ¿ the average number of defect per chip is directly proportional to the chip size. Initially, at t ¿ 0, the average number of defects per chip is relatively low, and the average value of yield exponentially decreases with the increase in chip size, and is in good agreement with Poisson´s yield model. After approximately 1 min of field application the average yield decreases with the increase in the time of field application, the chip size and the applied field; but increases with the increase in the oxide thickness. In general, the average yield decreases with the increase in the average number of defects per chip in a manner very close to Murphy´s yield curve, whatever the reasons for the increase in the average number of defects per chip.